< draft-ietf-tsvwg-aqm-dualq-coupled-08.txt   draft-ietf-tsvwg-aqm-dualq-coupled-09.txt >
Transport Area working group (tsvwg) K. De Schepper Transport Area working group (tsvwg) K. De Schepper
Internet-Draft Nokia Bell Labs Internet-Draft Nokia Bell Labs
Intended status: Experimental B. Briscoe, Ed. Intended status: Experimental B. Briscoe, Ed.
Expires: May 8, 2019 CableLabs Expires: January 6, 2020 G. White
O. Bondarenko CableLabs
Simula Research Lab July 05, 2019
I. Tsang
Nokia
November 04, 2018
DualQ Coupled AQMs for Low Latency, Low Loss and Scalable Throughput DualQ Coupled AQMs for Low Latency, Low Loss and Scalable Throughput
(L4S) (L4S)
draft-ietf-tsvwg-aqm-dualq-coupled-08 draft-ietf-tsvwg-aqm-dualq-coupled-09
Abstract Abstract
The Low Latency Low Loss Scalable Throughput (L4S) architecture The Low Latency Low Loss Scalable Throughput (L4S) architecture
allows data flows over the public Internet to predictably achieve allows data flows over the public Internet to predictably achieve
ultra-low queuing latency, generally zero congestion loss and scaling ultra-low queuing latency, generally zero congestion loss and scaling
of per-flow throughput without the problems of traditional TCP. To of per-flow throughput without the problems of traditional TCP. To
achieve this, L4S data flows use a 'scalable' congestion control achieve this, L4S data flows have to use one of the family of
similar to Data Centre TCP (DCTCP) and a form of Explicit Congestion 'Scalable' congestion controls (Data Centre TCP and TCP Prague are
Notification (ECN) with modified behaviour. However, until now, examples) and a form of Explicit Congestion Notification (ECN) with
scalable congestion controls did not co-exist with existing TCP Reno/ modified behaviour. However, until now, Scalable congestion controls
Cubic traffic---scalable controls are so aggressive that 'Classic' did not co-exist with existing TCP Reno/Cubic traffic --- Scalable
TCP algorithms drive themselves to starvation. Therefore, until now, controls are so aggressive that 'Classic' TCP algorithms drive
L4S controls could only be deployed where a clean-slate environment themselves to a small capacity share. Therefore, until now, L4S
could be arranged, such as in private data centres (hence the name controls could only be deployed where a clean-slate environment could
DCTCP). This specification defines `DualQ Coupled Active Queue be arranged, such as in private data centres (hence the name DCTCP).
Management (AQM)', which enables these scalable congestion controls This specification defines `DualQ Coupled Active Queue Management
to safely co-exist with Classic Internet traffic. (AQM)', which enables these Scalable congestion controls to safely
co-exist with Classic Internet traffic.
The Coupled AQM ensures that a flow runs at about the same rate The Coupled AQM ensures that competing Scalable and Classic flows run
whether it uses DCTCP or TCP Reno/Cubic. It achieves this at about the same rate. It achieves this indirectly, without having
indirectly, without having to inspect transport layer flow to inspect transport layer flow identifiers, When tested in a
identifiers, When tested in a residential broadband setting, DCTCP residential broadband setting, DCTCP also achieves sub-millisecond
also achieves sub-millisecond average queuing delay and zero average queuing delay and zero congestion loss under a wide range of
congestion loss under a wide range of mixes of DCTCP and `Classic' mixes of DCTCP and `Classic' broadband Internet traffic, without
broadband Internet traffic, without compromising the performance of compromising the performance of the Classic traffic. The solution
the Classic traffic. The solution also reduces network complexity also reduces network complexity and requires no configuration for the
and eliminates network configuration. public Internet.
Status of This Memo Status of This Memo
This Internet-Draft is submitted in full conformance with the This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79. provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet- working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/. Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
This Internet-Draft will expire on May 8, 2019. This Internet-Draft will expire on January 6, 2020.
Copyright Notice Copyright Notice
Copyright (c) 2018 IETF Trust and the persons identified as the Copyright (c) 2019 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of (https://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License. described in the Simplified BSD License.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1. Problem and Scope . . . . . . . . . . . . . . . . . . . . 3 1.1. Outline of the Problem . . . . . . . . . . . . . . . . . 3
1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 5 1.2. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3. Features . . . . . . . . . . . . . . . . . . . . . . . . 6 1.3. Terminology . . . . . . . . . . . . . . . . . . . . . . . 7
2. DualQ Coupled AQM . . . . . . . . . . . . . . . . . . . . . . 7 1.4. Features . . . . . . . . . . . . . . . . . . . . . . . . 8
2.1. Coupled AQM . . . . . . . . . . . . . . . . . . . . . . . 8 2. DualQ Coupled AQM . . . . . . . . . . . . . . . . . . . . . . 9
2.2. Dual Queue . . . . . . . . . . . . . . . . . . . . . . . 9 2.1. Coupled AQM . . . . . . . . . . . . . . . . . . . . . . . 9
2.3. Traffic Classification . . . . . . . . . . . . . . . . . 9 2.2. Dual Queue . . . . . . . . . . . . . . . . . . . . . . . 10
2.4. Overall DualQ Coupled AQM Structure . . . . . . . . . . . 10 2.3. Traffic Classification . . . . . . . . . . . . . . . . . 11
2.5. Normative Requirements for a DualQ Coupled AQM . . . . . 12 2.4. Overall DualQ Coupled AQM Structure . . . . . . . . . . . 11
2.5.1. Functional Requirements . . . . . . . . . . . . . . . 12 2.5. Normative Requirements for a DualQ Coupled AQM . . . . . 14
2.5.1.1. Requirements in Unexpected Cases . . . . . . . . 13 2.5.1. Functional Requirements . . . . . . . . . . . . . . . 14
2.5.2. Management Requirements . . . . . . . . . . . . . . . 15 2.5.1.1. Requirements in Unexpected Cases . . . . . . . . 15
3. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16 2.5.2. Management Requirements . . . . . . . . . . . . . . . 16
4. Security Considerations . . . . . . . . . . . . . . . . . . . 16 2.5.2.1. Configuration . . . . . . . . . . . . . . . . . . 16
4.1. Overload Handling . . . . . . . . . . . . . . . . . . . . 16 2.5.2.2. Monitoring . . . . . . . . . . . . . . . . . . . 18
2.5.2.3. Anomaly Detection . . . . . . . . . . . . . . . . 18
2.5.2.4. Deployment, Coexistence and Scaling . . . . . . . 19
3. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 19
4. Security Considerations . . . . . . . . . . . . . . . . . . . 19
4.1. Overload Handling . . . . . . . . . . . . . . . . . . . . 19
4.1.1. Avoiding Classic Starvation: Sacrifice L4S Throughput 4.1.1. Avoiding Classic Starvation: Sacrifice L4S Throughput
or Delay? . . . . . . . . . . . . . . . . . . . . . . 17 or Delay? . . . . . . . . . . . . . . . . . . . . . . 20
4.1.2. Congestion Signal Saturation: Introduce L4S Drop or 4.1.2. Congestion Signal Saturation: Introduce L4S Drop or
Delay? . . . . . . . . . . . . . . . . . . . . . . . 18 Delay? . . . . . . . . . . . . . . . . . . . . . . . 21
4.1.3. Protecting against Unresponsive ECN-Capable Traffic . 19 4.1.3. Protecting against Unresponsive ECN-Capable Traffic . 22
5. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 19 5. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 22
6. References . . . . . . . . . . . . . . . . . . . . . . . . . 20 6. Contributors . . . . . . . . . . . . . . . . . . . . . . . . 23
6.1. Normative References . . . . . . . . . . . . . . . . . . 20 7. References . . . . . . . . . . . . . . . . . . . . . . . . . 23
6.2. Informative References . . . . . . . . . . . . . . . . . 20 7.1. Normative References . . . . . . . . . . . . . . . . . . 23
Appendix A. Example DualQ Coupled PI2 Algorithm . . . . . . . . 23 7.2. Informative References . . . . . . . . . . . . . . . . . 24
A.1. Pass #1: Core Concepts . . . . . . . . . . . . . . . . . 23 Appendix A. Example DualQ Coupled PI2 Algorithm . . . . . . . . 27
A.2. Pass #2: Overload Details . . . . . . . . . . . . . . . . 30 A.1. Pass #1: Core Concepts . . . . . . . . . . . . . . . . . 28
Appendix B. Example DualQ Coupled Curvy RED Algorithm . . . . . 33 A.2. Pass #2: Overload Details . . . . . . . . . . . . . . . . 36
Appendix C. Guidance on Controlling Throughput Equivalence . . . 39 Appendix B. Example DualQ Coupled Curvy RED Algorithm . . . . . 40
Appendix D. Open Issues . . . . . . . . . . . . . . . . . . . . 40 B.1. Curvy RED in Pseudocode . . . . . . . . . . . . . . . . . 40
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 41 B.2. Efficient Implementation of Curvy RED . . . . . . . . . . 46
Appendix C. Guidance on Controlling Throughput Equivalence . . . 48
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 49
1. Introduction 1. Introduction
1.1. Problem and Scope This document specifies a framework for DualQ Coupled AQMs, which is
the network part of the L4S architecture [I-D.ietf-tsvwg-l4s-arch].
L4S enables both ultra-low queuing latency and high throughput at the
same time, for ad hoc numbers of capacity-seeking applications all
sharing the same capacity.
1.1. Outline of the Problem
Latency is becoming the critical performance factor for many (most?) Latency is becoming the critical performance factor for many (most?)
applications on the public Internet, e.g. interactive Web, Web applications on the public Internet, e.g. interactive Web, Web
services, voice, conversational video, interactive video, interactive services, voice, conversational video, interactive video, interactive
remote presence, instant messaging, online gaming, remote desktop, remote presence, instant messaging, online gaming, remote desktop,
cloud-based applications, and video-assisted remote control of cloud-based applications, and video-assisted remote control of
machinery and industrial processes. In the developed world, further machinery and industrial processes. In the developed world, further
increases in access network bit-rate offer diminishing returns, increases in access network bit-rate offer diminishing returns,
whereas latency is still a multi-faceted problem. In the last decade whereas latency is still a multi-faceted problem. In the last decade
or so, much has been done to reduce propagation time by placing or so, much has been done to reduce propagation time by placing
caches or servers closer to users. However, queuing remains a major caches or servers closer to users. However, queuing remains a major
intermittent component of latency. intermittent component of latency.
The Diffserv architecture provides Expedited Forwarding [RFC3246], so Traditionally ultra-low latency has only been available for a few
that low latency traffic can jump the queue of other traffic. selected low rate applications, that confine their sending rate
However, on access links dedicated to individual sites (homes, small within a specially carved-off portion of capacity, which is
enterprises or mobile devices), often all traffic at any one time prioritized over other traffic, e.g. Diffserv EF [RFC3246]. Up to
will be latency-sensitive and, if all the traffic on a link is marked now it has not been possible to allow any number of low latency, high
as EF, Diffserv cannot reduce the delay of any of it. In contrast, throughput applications to seek to fully utilize available capacity,
the Low Latency Low Loss Scalable throughput (L4S) approach removes because the capacity-seeking process itself causes too much queuing
the causes of any unnecessary queuing delay. delay.
The bufferbloat project has shown that excessively-large buffering To reduce this queuing delay caused by the capacity seeking process,
(`bufferbloat') has been introducing significantly more delay than changes either to the network alone or to end-systems alone are in
the underlying propagation time. These delays appear only progress. L4S involves a recognition that both approaches are
intermittently--only when a capacity-seeking (e.g. TCP) flow is long yielding diminishing returns:
enough for the queue to fill the buffer, making every packet in other
flows sharing the buffer sit through the queue.
Active queue management (AQM) was originally developed to solve this o Recent state-of-the-art active queue management (AQM) in the
problem (and others). Unlike Diffserv, which gives low latency to network, e.g. fq_CoDel [RFC8290], PIE [RFC8033], Adaptive
some traffic at the expense of others, AQM controls latency for _all_ RED [ARED01] ) has reduced queuing delay for all traffic, not just
traffic in a class. In general, AQMs introduce an increasing level a select few applications. However, no matter how good the AQM,
of discard from the buffer the longer the queue persists above a the capacity-seeking (sawtoothing) rate of TCP-like congestion
shallow threshold. This gives sufficient signals to capacity-seeking controls represents a lower limit that will either cause queuing
(aka. greedy) flows to keep the buffer empty for its intended delay to vary or cause the link to be under-utilized. These AQMs
purpose: absorbing bursts. However, RED [RFC2309] and other are tuned to allow a typical capacity-seeking TCP-Friendly flow to
algorithms from the 1990s were sensitive to their configuration and induce an average queue that roughly doubles the base RTT, adding
hard to set correctly. So, AQM was not widely deployed in the 1990s. 5-15 ms of queuing on average (cf. 500 microseconds with L4S for
the same mix of long-running and web traffic). However, for many
applications low delay is not useful unless it is consistently
low. With these AQMs, 99th percentile queuing delay is 20-30 ms
(cf. 2 ms with the same traffic over L4S).
More recent state-of-the-art AQMs, e.g. fq_CoDel [RFC8290], o Similarly, recent research into using e2e congestion control
PIE [RFC8033], Adaptive RED [ARED01], are easier to configure, without needing an AQM in the network (e.g.BBRv1 [BBRv1]) seems to
because they define the queuing threshold in time not bytes, so it is have hit a similar lower limit to queuing delay of about 20ms on
invariant for different link rates. However, no matter how good the average (and any additional BBRv1 flow adds another 20ms of
AQM, the sawtoothing rate of TCP will either cause queuing delay to queuing) but there are also regular 25ms delay spikes due to
vary or cause the link to be under-utilized. Even with a perfectly bandwidth probes and 60ms spikes due to flow-starts.
tuned AQM, the additional queuing delay will be of the same order as
the underlying speed-of-light delay across the network. Flow-queuing
can isolate one flow from another, but it cannot isolate a TCP flow
from the delay variations it inflicts on itself, and it has other
problems - it overrides the flow rate decisions of variable rate
video applications, it does not recognise the flows within IPSec VPN
tunnels and it is relatively expensive to implement.
It seems that further changes to the network alone will now yield L4S learns from the experience of Data Center TCP [RFC8257], which
diminishing returns. Data Centre TCP (DCTCP [RFC8257]) teaches us shows the power of complementary changes both in the network and on
that a small but radical change to TCP is needed to cut two major end-systems. DCTCP teaches us that two small but radical changes to
outstanding causes of queuing delay variability: congestion control are needed to cut the two major outstanding causes
of queuing delay variability:
1. the `sawtooth' varying rate of TCP itself; 1. Far smaller rate variations (sawteeth) than TCP-Friendly
congestion controls;
2. the smoothing delay deliberately introduced into AQMs to permit 2. A shift of smoothing and hence smoothing delay from network to
bursts without triggering losses. sender.
The former causes a flow's round trip time (RTT) to vary from about 1 Without the former, a 'Classic' flow's round trip time (RTT) varies
to 2 times the base RTT between the machines in question. The latter between roughly 1 and 2 times the base RTT between the machines in
delays the system's response to change by a worst-case question. Without the latter a 'Classic' flow's response to changing
(transcontinental) RTT, which could be hundreds of times the actual events is delayed by a worst-case (transcontinental) RTT, which could
RTT of typical traffic from localized CDNs. be hundreds of times the actual smoothing delay needed for the RTT of
typical traffic from localized CDNs.
Latency is not our only concern: These changes are the two main features of the family of so-called
'Scalable' congestion controls (which includes DCTCP). Both these
changes only reduce delay in combination with a complementary change
in the network and they are both only feasible with ECN, not drop,
for the signalling:
3. It was known when TCP was first developed that it would not scale 1. The smaller sawteeth need an extremely shallow ECN packet-marking
to high bandwidth-delay products [TCP-CA]. threshold in the queue.
Given regular broadband bit-rates over WAN distances are 2. And no smoothing in the network means that every fluctuation of
already [RFC3649] beyond the scaling range of `classic' TCP Reno, the queue is signalled immediately.
`less unscalable' Cubic [RFC8312] and
Compound [I-D.sridharan-tcpm-ctcp] variants of TCP have been Without ECN, either of these would lead to very high loss levels.
successfully deployed. However, these are now approaching their But, with ECN, the resulting high marking levels are fine.
scaling limits. Unfortunately, fully scalable TCPs such as DCTCP
cause `classic' TCP to starve itself, which is why they have been However, until now, Scalable congestion controls (like DCTCP) did not
confined to private data centres or research testbeds (until now). co-exist with existing ECN-capable TCP Reno [RFC5681] or Cubic
[RFC8312] traffic --- Scalable controls are so aggressive that these
'Classic' TCP algorithms drive themselves to a small capacity share.
Therefore, until now, L4S controls could only be deployed where a
clean-slate environment could be arranged, such as in private data
centres (hence the name DCTCP).
This document specifies a `DualQ Coupled AQM' extension that solves This document specifies a `DualQ Coupled AQM' extension that solves
the problem of coexistence between scalable and classic flows, the problem of coexistence between Scalable and Classic flows,
without having to inspect flow identifiers. The AQM is not like without having to inspect flow identifiers. It is not like flow-
flow-queuing approaches [RFC8290] that classify packets by flow queuing approaches [RFC8290] that classify packets by flow identifier
identifier into numerous separate queues in order to isolate sparse into separate queues in order to isolate sparse flows from the higher
flows from the higher latency in the queues assigned to heavier latency in the queues assigned to heavier flows. If a flow needs
flows. In contrast, the AQM exploits the behaviour of scalable both low delay and high throughput, having a queue to itself does not
congestion controls like DCTCP so that every packet in every flow isolate it from the harm it causes to itself. In contrast, L4S
sharing the queue for DCTCP-like traffic can be served with very low addresses the root cause of the latency problem --- it is an enabler
latency. for the smooth low latency scalable behaviour of Scalable congestion
controls, so that every packet in every flow can enjoy very low
latency, then there is no need to isolate each flow into a separate
queue.
This AQM extension can be combined with any AQM designed for a single 1.2. Scope
queue that generates a statistical or deterministic mark/drop
probability driven by the queue dynamics. In many cases it L4S involves complementary changes in the network and on end-systems:
simplifies the basic control algorithm, and requires little extra
processing. Therefore it is believed the Coupled AQM would be Network: A DualQ Coupled AQM (defined in the present document);
applicable and easy to deploy in all types of buffers; buffers in
cost-reduced mass-market residential equipment; buffers in end-system End-system: A Scalable congestion control (defined in Section 2.1.
stacks; buffers in carrier-scale equipment including remote access
servers, routers, firewalls and Ethernet switches; buffers in network Packet identifier: The network and end-system parts of L4S can be
interface cards, buffers in virtualized network appliances, deployed incrementally, because they both identify L4S packets
hypervisors, and so on. using the experimentally assigned explicit congestion notification
(ECN) codepoints in the IP header: ECT(1) and CE [RFC8311]
[I-D.ietf-tsvwg-ecn-l4s-id].
Data Center TCP (DCTCP [RFC8257]) is an example of a Scalable
congestion control that has been deployed for some time in Linux,
Windows and FreeBSD operating systems and Relentless TCP [Mathis09]
is another example. During the progress of this document through the
IETF a number of other Scalable congestion controls were implemented,
e.g. TCP Prague [PragueLinux], QUIC Prague and the L4S variant of
SCREAM for real-time media [RFC8298]. (Note: after the v3.19 Linux
kernel, bugs were introduced into DCTCP's scalable behaviour and not
all the patches applied for L4S evaluation had been applied to the
mainline Linux kernel, which was at v5.2 at the time of writing).
The focus of this specification is to get the network part of the L4S
service in place. Then, without any management intervention,
applications can exploit this new network capability as their
operating systems migrate to Scalable congestion controls, which can
then evolve _while_ their benefits are being enjoyed by everyone on
the Internet.
The DualQ Coupled AQM framework can incorporate any AQM designed for
a single queue that generates a statistical or deterministic mark/
drop probability driven by the queue dynamics. Pseudocode examples
of two different DualQ Coupled AQMs are given the appendices. In
many cases the framework simplifies the basic control algorithm, and
requires little extra processing. Therefore it is believed the
Coupled AQM would be applicable and easy to deploy in all types of
buffers; buffers in cost-reduced mass-market residential equipment;
buffers in end-system stacks; buffers in carrier-scale equipment
including remote access servers, routers, firewalls and Ethernet
switches; buffers in network interface cards, buffers in virtualized
network appliances, hypervisors, and so on.
For the public Internet, nearly all the benefit will typically be For the public Internet, nearly all the benefit will typically be
achieved by deploying the Coupled AQM into either end of the access achieved by deploying the Coupled AQM into either end of the access
link between a 'site' and the Internet, which is invariably the link between a 'site' and the Internet, which is invariably the
bottleneck. Here, the term 'site' is used loosely to mean a home, an bottleneck. Here, the term 'site' is used loosely to mean a home, an
office, a campus or mobile user equipment. office, a campus or mobile user equipment.
Latency is not the only concern of L4S:
o The 'Low Loss" part of the name denotes that L4S generally
achieves zero congestion loss (which would otherwise cause
retransmission delays), due to its use of ECN.
o The "Scalable throughput" part of the name denotes that the per-
flow throughput of Scalable congestion controls should scale
indefinitely, avoiding the imminent scaling problems with TCP-
Friendly congestion control algorithms [RFC3649].
The former is clearly in scope of this AQM document. However, the
latter is an outcome of the end-system behaviour, and therefore
outside the scope of this AQM document, even though the AQM is an
enabler.
The overall L4S architecture [I-D.ietf-tsvwg-l4s-arch] gives more The overall L4S architecture [I-D.ietf-tsvwg-l4s-arch] gives more
detail, including on wider deployment aspects such as coexistence in detail, including on wider deployment aspects such as backwards
bottlenecks where a DualQ Coupled AQM has not been deployed. The compatibility of Scalable congestion controls in bottlenecks where a
supporting papers [PI2] and [DCttH15] give the full rationale for the DualQ Coupled AQM has not been deployed. The supporting papers [PI2]
AQM's design, both discursively and in more precise mathematical and [DCttH15] give the full rationale for the AQM's design, both
form. discursively and in more precise mathematical form.
1.2. Terminology 1.3. Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in [RFC2119] when, and document are to be interpreted as described in [RFC2119] when, and
only when, they appear in all capitals, as shown here. only when, they appear in all capitals, as shown here.
The DualQ Coupled AQM uses two queues for two services. Each of the The DualQ Coupled AQM uses two queues for two services. Each of the
following terms identifies both the service and the queue that following terms identifies both the service and the queue that
provides the service: provides the service:
Classic (denoted by subscript C): The `Classic' service is intended Classic (denoted by subscript C): The `Classic' service is intended
for all the behaviours that currently co-exist with TCP Reno (TCP for all the behaviours that currently co-exist with TCP Reno (TCP
Cubic, Compound, SCTP, etc). Cubic, Compound, SCTP, etc).
Low-Latency, Low-Loss and Scalable (L4S, denoted by subscript L): Low-Latency, Low-Loss and Scalable (L4S, denoted by subscript L):
The `L4S' service is intended for a set of congestion controls The `L4S' service is intended for a set of congestion controls
with scalable properties (e.g. DCTCP [RFC8257], Relentless with scalable properties, such as TCP Prague and DCTCP. For the
TCP [Mathis09], the L4S variant of SCREAM for real-time public Internet an L4S transport has to comply with the
media {ToDo: ref}). For the public Internet a scalable control requirements in Section 4 of [I-D.ietf-tsvwg-ecn-l4s-id] (aka.
has to comply with the requirements in [I-D.ietf-tsvwg-ecn-l4s-id] the 'Prague L4S requirements').
(aka. the 'TCP Prague requirements').
Either service can cope with a proportion of unresponsive or less- Either service can cope with a proportion of unresponsive or less-
responsive traffic as well, as long (e.g. DNS, VoIP, game sync responsive traffic as well, as long (e.g. DNS, VoIP, game sync
datagrams, etc), just as a single queue AQM can if this traffic makes datagrams, etc), just as a single queue AQM can if this traffic makes
minimal contribution to queuing. The DualQ Coupled AQM behaviour minimal contribution to queuing. The DualQ Coupled AQM behaviour
below is defined to be similar to a single FIFO queue with respect to below is defined to be similar to a single FIFO queue with respect to
unresponsive and overload traffic. unresponsive and overload traffic.
1.3. Features 1.4. Features
The AQM couples marking and/or dropping across the two queues such The AQM couples marking and/or dropping from the Classic queue to the
that a flow will get roughly the same throughput whichever it uses. L4S queue in such a way that a flow will get roughly the same
Therefore both queues can feed into the full capacity of a link and throughput whichever it uses. Therefore both queues can feed into
no rates need to be configured for the queues. The L4S queue enables the full capacity of a link and no rates need to be configured for
scalable congestion controls like DCTCP to give stunningly low and the queues. The L4S queue enables Scalable congestion controls like
predictably low latency, without compromising the performance of DCTCP or TCP Prague to give stunningly low and predictably low
competing 'Classic' Internet traffic. Thousands of tests have been latency, without compromising the performance of competing 'Classic'
conducted in a typical fixed residential broadband setting. Typical Internet traffic.
experiments used base round trip delays up to 100ms between the data
centre and home network, and large amounts of background traffic in Thousands of tests have been conducted in a typical fixed residential
both queues. For every L4S packet, the AQM kept the average queuing broadband setting. Experiments used a range of base round trip
delay below 1ms (or 2 packets if serialization delay is bigger for delays up to 100ms and link rates up to 200 Mb/s between the data
slow links), and no losses at all were introduced by the AQM. centre and home network, with varying amounts of background traffic
Details of the extensive experiments are available [PI2] [DCttH15]. in both queues. For every L4S packet, the AQM kept the average
queuing delay below 1ms (or 2 packets where serialization delay
exceeded 1ms on slower links), with 99th percentile no worse than
2ms. No losses at all were introduced by the L4S AQM. Details of
the extensive experiments are available [PI2] [DCttH15].
Subjective testing was also conducted by multiple people all Subjective testing was also conducted by multiple people all
simultaneously using very demanding high bandwidth low latency simultaneously using very demanding high bandwidth low latency
applications over a single shared access link [L4Sdemo16]. In one applications over a single shared access link [L4Sdemo16]. In one
application, each user could use finger gestures to pan or zoom their application, each user could use finger gestures to pan or zoom their
own high definition (HD) sub-window of a larger video scene generated own high definition (HD) sub-window of a larger video scene generated
on the fly in 'the cloud' from a football match. Another user on the fly in 'the cloud' from a football match. Another user
wearing VR goggles was remotely receiving a feed from a 360-degree wearing VR goggles was remotely receiving a feed from a 360-degree
camera in a racing car, again with the sub-window in their field of camera in a racing car, again with the sub-window in their field of
vision generated on the fly in 'the cloud' dependent on their head vision generated on the fly in 'the cloud' dependent on their head
skipping to change at page 7, line 18 skipping to change at page 8, line 48
latency was so low that the football picture appeared to stick to the latency was so low that the football picture appeared to stick to the
user's finger on the touchpad and the experience fed from the remote user's finger on the touchpad and the experience fed from the remote
camera did not noticeably lag head movements. All the L4S data (even camera did not noticeably lag head movements. All the L4S data (even
including the downloads) achieved the same ultra-low latency. With including the downloads) achieved the same ultra-low latency. With
an alternative AQM, the video noticeably lagged behind the finger an alternative AQM, the video noticeably lagged behind the finger
gestures and head movements. gestures and head movements.
Unlike Diffserv Expedited Forwarding, the L4S queue does not have to Unlike Diffserv Expedited Forwarding, the L4S queue does not have to
be limited to a small proportion of the link capacity in order to be limited to a small proportion of the link capacity in order to
achieve low delay. The L4S queue can be filled with a heavy load of achieve low delay. The L4S queue can be filled with a heavy load of
capacity-seeking flows like DCTCP and still achieve low delay. The capacity-seeking flows (TCP Prague etc.) and still achieve low delay.
L4S queue does not rely on the presence of other traffic in the The L4S queue does not rely on the presence of other traffic in the
Classic queue that can be 'overtaken'. It gives low latency to L4S Classic queue that can be 'overtaken'. It gives low latency to L4S
traffic whether or not there is Classic traffic, and the latency of traffic whether or not there is Classic traffic, and the latency of
Classic traffic does not suffer when a proportion of the traffic is Classic traffic does not suffer when a proportion of the traffic is
L4S. The two queues are only necessary because DCTCP-like flows L4S.
cannot keep latency predictably low and keep utilization high if they
are mixed with legacy TCP flows,
The experiments used the Linux implementation of DCTCP that is The two queues are only necessary because:
deployed in private data centres, without any modification despite
its known deficiencies. Nonetheless, certain modifications will be o the large variations (sawteeth) of Classic flows need roughly a
necessary before DCTCP is safe to use on the Internet, which are base RTT of queuing delay to ensure full utilization
recorded in Appendix A of [I-D.ietf-tsvwg-ecn-l4s-id]. However, the
focus of this specification is to get the network service in place. o while Scalable flows do not need a queue to keep utilization high,
Then, without any management intervention, applications can exploit but they cannot keep latency predictably low if they are mixed
it by migrating to scalable controls like DCTCP, which can then with legacy TCP flows,
evolve _while_ their benefits are being enjoyed by everyone on the
Internet. The L4S queue has latency priority, but the coupling from the Classic
to the L4S AQM (explained below) ensures that it does not have
bandwidth priority over the Classic queue.
2. DualQ Coupled AQM 2. DualQ Coupled AQM
There are two main aspects to the approach: There are two main aspects to the approach:
o the Coupled AQM that addresses throughput equivalence between o the Coupled AQM that addresses throughput equivalence between
Classic (e.g. Reno, Cubic) flows and L4S flows (that satisfy the Classic (e.g. Reno, Cubic) flows and L4S flows (that satisfy the
TCP Prague requirements). Prague L4S requirements).
o the Dual Queue structure that provides latency separation for L4S o the Dual Queue structure that provides latency separation for L4S
flows to isolate them from the typically large Classic queue. flows to isolate them from the typically large Classic queue.
2.1. Coupled AQM 2.1. Coupled AQM
In the 1990s, the `TCP formula' was derived for the relationship In the 1990s, the `TCP formula' was derived for the relationship
between TCP's congestion window, cwnd, and its drop probability, p. between TCP's congestion window, cwnd, and its drop probability, p.
To a first order approximation, cwnd of TCP Reno is inversely To a first order approximation, cwnd of TCP Reno is inversely
proportional to the square root of p. proportional to the square root of p.
We focus on Reno as the worst case, because if we do not harm Reno, The design focuses on Reno as the worst case, because if it does no
we will not harm Cubic. Nonetheless, TCP Cubic implements a Reno- harm to Reno, it will not harm Cubic or any traffic designed to be
compatibility mode, which is the only relevant mode for typical RTTs friendly to Reno. TCP Cubic implements a Reno-compatibility mode,
under 20ms as long as the throughput of a single flow is less than which is relevant for typical RTTs under 20ms as long as the
about 500Mb/s. Therefore it can be assumed that Cubic traffic throughput of a single flow is less than about 700Mb/s. In such
behaves similarly to Reno (but with a slightly different constant of cases it can be assumed that Cubic traffic behaves similarly to Reno
proportionality). The term 'Classic' will be used for the collection (but with a slightly different constant of proportionality). The
of Reno-friendly traffic including Cubic in Reno mode. term 'Classic' will be used for the collection of Reno-friendly
traffic including Cubic in Reno mode.
The supporting paper [PI2] includes the derivation of the equivalent The supporting paper [PI2] includes the derivation of the equivalent
rate equation for DCTCP, for which cwnd is inversely proportional to rate equation for DCTCP, for which cwnd is inversely proportional to
p (not the square root), where in this case p is the ECN marking p (not the square root), where in this case p is the ECN marking
probability. DCTCP is not the only congestion control that behaves probability. DCTCP is not the only congestion control that behaves
like this, so the term 'L4S' traffic will be used for all similar like this, so the term 'Scalable' will be used for all similar
behaviour. congestion control behaviours (see examples in Section 1.2). The
term 'L4S' is also used for traffic driven by a Scalable congestion
control that also complies with the additional 'Prague L4S'
requirements [I-D.ietf-tsvwg-ecn-l4s-id].
For safe co-existence, under stationary conditions, a DCTCP flow has For safe co-existence, under stationary conditions, a Scalable flow
to run at roughly the same rate as a Reno TCP flow (all other factors has to run at roughly the same rate as a Reno TCP flow (all other
being equal). So the drop or marking probability for Classic factors being equal). So the drop or marking probability for Classic
traffic, p_C has to be distinct from the marking probability for L4S traffic, p_C has to be distinct from the marking probability for L4S
traffic, p_L. [RFC8311] updates the original ECN specification traffic, p_L. [RFC8311] updates the original ECN specification
[RFC3168] to allow these probabilities to be distinct, because RFC [RFC3168] to allow these probabilities to be distinct, because RFC
3168 required them to be the same. 3168 required them to be the same.
Also, to remain stable, Classic sources need the network to smooth Also, to remain stable, Classic sources need the network to smooth
p_C so it changes relatively slowly. In contrast, L4S avoids p_C so it changes relatively slowly. It is hard for a network node
smoothing in the network, because it delays all signals for a worst- to know the RTTs of all the flows, so a Classic AQM adds a _worst-
case RTT. So instead, L4S sources smooth the ECN marking probability case_ RTT of smoothing delay (about 100-200 ms). In contrast, L4S
themselves, so they expect the network to generate ECN marks with a shifts responsibility for smoothing ECN feedback to the sender, which
probability p_L that tracks the instantaneous unsmoothed queue. only delays its response by its _own_ RTT, and allows a more
immediate response if necessary.
The Coupled AQM achieves safe coexistence by making the Classic drop The Coupled AQM achieves safe coexistence by making the Classic drop
probability p_C proportional to the square of the coupled L4S probability p_C proportional to the square of the coupled L4S
probability p_CL. p_CL is an input to the instantaneous L4S marking probability p_CL. p_CL is an input to the instantaneous L4S marking
probability p_L but it changes as slowly as p_C. This makes the Reno probability p_L but it changes as slowly as p_C. This makes the Reno
flow rate roughly equal the DCTCP flow rate, because the squaring of flow rate roughly equal the DCTCP flow rate, because the squaring of
p_CL counterbalances the square root of p_C in the Classic 'TCP p_CL counterbalances the square root of p_C in the Classic 'TCP
formula'. formula'.
Stating this as a formula, the relation between Classic drop Stating this as a formula, the relation between Classic drop
probability, p_C, and the coupled L4S probability p_CL needs to take probability, p_C, and the coupled L4S probability p_CL needs to take
the form: the form:
p_C = ( p_CL / k )^2 (1) p_C = ( p_CL / k )^2 (1)
where k is the constant of proportionality, which we shall call the where k is the constant of proportionality, which is termed the
coupling factor. coupling factor.
2.2. Dual Queue 2.2. Dual Queue
Classic traffic typically builds a large queue to prevent under- Classic traffic needs to build a large queue to prevent under-
utilization. Therefore a separate queue is provided for L4S traffic, utilization. Therefore a separate queue is provided for L4S traffic,
and it is scheduled with priority over Classic. Priority is and it is scheduled with priority over the Classic queue. Priority
conditional to prevent starvation of Classic traffic. is conditional to prevent starvation of Classic traffic.
Nonetheless, coupled marking ensures that giving priority to L4S Nonetheless, coupled marking ensures that giving priority to L4S
traffic still leaves the right amount of spare scheduling time for traffic still leaves the right amount of spare scheduling time for
Classic flows to each get equivalent throughput to DCTCP flows (all Classic flows to each get equivalent throughput to DCTCP flows (all
other factors such as RTT being equal). other factors such as RTT being equal).
2.3. Traffic Classification 2.3. Traffic Classification
Both the Coupled AQM and DualQ mechanisms need an identifier to Both the Coupled AQM and DualQ mechanisms need an identifier to
distinguish L and C packets. Then the coupling algorithm can achieve distinguish L and C packets. Then the coupling algorithm can achieve
coexistence without having to inspect flow identifiers, because it coexistence without having to inspect flow identifiers, because it
can apply the appropriate marking or dropping probability to all can apply the appropriate marking or dropping probability to all
flows of each type. A separate flows of each type. A separate
specification [I-D.ietf-tsvwg-ecn-l4s-id] requires the sender to use specification [I-D.ietf-tsvwg-ecn-l4s-id] requires the sender to use
the ECT(1) codepoint of the ECN field as this identifier, having the ECT(1) and CE codepoints of the ECN field as this identifier,
assessed various alternatives. An additional process document has having assessed various alternatives. An additional process document
proved necessary to make the ECT(1) codepoint available for has proved necessary to make the ECT(1) codepoint available for
experimentation [RFC8311]. experimentation [RFC8311].
For policy reasons, an operator might choose to steer certain packets For policy reasons, an operator might choose to steer certain packets
(e.g. from certain flows or with certain addresses) out of the L (e.g. from certain flows or with certain addresses) out of the L
queue, even though they identify themselves as L4S by their ECN queue, even though they identify themselves as L4S by their ECN
codepoints. In such cases, the device MUST NOT alter the ECN field, codepoints. In such cases, [I-D.ietf-tsvwg-ecn-l4s-id] says that the
so that it is preserved end-to-end. The aim is that each operator device "MUST NOT alter the end-to-end L4S ECN identifier", so that it
can choose how it treats L4S traffic locally, but an individual is preserved end-to-end. The aim is that each operator can choose
operator does not alter the identification of L4S packets, which how it treats L4S traffic locally, but an individual operator does
would prevent other operators downstream from making their own not alter the identification of L4S packets, which would prevent
choices on how to treat L4S traffic. other operators downstream from making their own choices on how to
treat L4S traffic.
In addition, other identifiers could be used to classify certain In addition, an operator could use other identifiers to classify
additional packet types into the L queue, that are deemed not to risk certain additional packet types into the L queue that it deems will
harming the L4S service. For instance addresses of specific not risk harm to the L4S service. For instance addresses of specific
applications or hosts (see [I-D.ietf-tsvwg-ecn-l4s-id]), specific applications or hosts (see [I-D.ietf-tsvwg-ecn-l4s-id]), specific
Diffserv codepoints such as EF (Expedited Forwarding) and Voice-Admit Diffserv codepoints such as EF (Expedited Forwarding) and Voice-Admit
service classes (see [I-D.briscoe-tsvwg-l4s-diffserv]) or certain service classes (see [I-D.briscoe-tsvwg-l4s-diffserv]) or certain
protocols (e.g. ARP, DNS). protocols (e.g. ARP, DNS). Note that the mechanism only reads these
identifiers. [I-D.ietf-tsvwg-ecn-l4s-id] says it "MUST NOT alter
Note that the mechanism only reads these classifiers, it MUST NOT re- these non-ECN identifiers".
mark or alter these identifiers (except for marking the ECN field
with the CE codepoint - with increasing frequency to indicate
increasing congestion).
2.4. Overall DualQ Coupled AQM Structure 2.4. Overall DualQ Coupled AQM Structure
Figure 1 shows the overall structure that any DualQ Coupled AQM is Figure 1 shows the overall structure that any DualQ Coupled AQM is
likely to have. This schematic is intended to aid understanding of likely to have. This schematic is intended to aid understanding of
the current designs of DualQ Coupled AQMs. However, it is not the current designs of DualQ Coupled AQMs. However, it is not
intended to preclude other innovative ways of satisfying the intended to preclude other innovative ways of satisfying the
normative requirements in Section 2.5 that minimally define a DualQ normative requirements in Section 2.5 that minimally define a DualQ
Coupled AQM. Coupled AQM.
skipping to change at page 10, line 49 skipping to change at page 12, line 29
So the slow-moving input to ECN marking in the L queue (the coupled So the slow-moving input to ECN marking in the L queue (the coupled
L4S probability) is: L4S probability) is:
p_CL = k*p', (3) p_CL = k*p', (3)
where k is the constant coupling factor (see Appendix C). where k is the constant coupling factor (see Appendix C).
It can be seen that these two transformations of p' implement the It can be seen that these two transformations of p' implement the
required coupling given in equation (1) earlier. required coupling given in equation (1) earlier.
The actual probability p_L that we apply to the L queue needs to The actual ECN marking probability p_L that is applied to the L queue
track the immediate L queue delay, as well as track p_CL under needs to track the immediate L queue delay under L-only congestion
stationary conditions. So we use a native AQM in the L queue that conditions, as well as track p_CL under coupled congestion
calculates a probability p'_L as a function of the instantaneous L conditions. So the L queue uses a native AQM that calculates a
queue. And, given the L queue has conditional strict priority over probability p'_L as a function of the instantaneous L queue delay.
the C queue, whenever the L queue grows, we should apply marking And, given the L queue has conditional strict priority over the C
queue, whenever the L queue grows, the AQM should apply marking
probability p'_L, but p_L should not fall below p_CL. This suggests: probability p'_L, but p_L should not fall below p_CL. This suggests:
p_L = max(p'_L, p_CL), (4) p_L = max(p'_L, p_CL), (4)
which has also been found to work very well in practice. which has also been found to work very well in practice.
_________ _________
| | ,------. | | ,------.
L4S queue | |===>| ECN | L4S queue | |===>| ECN |
,'| _______|_| |marker|\ ,'| _______|_| |marker|\
skipping to change at page 12, line 8 skipping to change at page 13, line 47
where a continually busy L4S queue blocks a DNS request in the where a continually busy L4S queue blocks a DNS request in the
Classic queue, arbitrarily delaying the start of a new Classic flow. Classic queue, arbitrarily delaying the start of a new Classic flow.
Example DualQ Coupled AQM algorithms called DualPI2 and Curvy RED are Example DualQ Coupled AQM algorithms called DualPI2 and Curvy RED are
given in Appendix A and Appendix B. Either example AQM can be used given in Appendix A and Appendix B. Either example AQM can be used
to couple packet marking and dropping across a dual Q. to couple packet marking and dropping across a dual Q.
DualPI2 uses a Proportional-Integral (PI) controller as the Base AQM. DualPI2 uses a Proportional-Integral (PI) controller as the Base AQM.
Indeed, this Base AQM with just the squared output and no L4S queue Indeed, this Base AQM with just the squared output and no L4S queue
can be used as a drop-in replacement for PIE [RFC8033], in which case can be used as a drop-in replacement for PIE [RFC8033], in which case
we call it just PI2 [PI2]. PI2 is a principled simplification of PIE it is just called PI2 [PI2]. PI2 is a principled simplification of
that is both more responsive and more stable in the face of PIE that is both more responsive and more stable in the face of
dynamically varying load. dynamically varying load.
Curvy RED is derived from RED [RFC2309], but its configuration Curvy RED is derived from RED [RFC2309], but its configuration
parameters are insensitive to link rate and it requires less parameters are insensitive to link rate and it requires less
operations per packet. However, DualPI2 is more responsive and operations per packet. However, DualPI2 is more responsive and
stable over a wider range of RTTs than Curvy RED. As a consequence, stable over a wider range of RTTs than Curvy RED. As a consequence,
DualPI2 has attracted more development attention than Curvy RED, DualPI2 has attracted more development and evaluation attention than
leaving the Curvy RED design incomplete and not so fully evaluated. Curvy RED, leaving the Curvy RED design incomplete and not so fully
evaluated.
Both AQMs regulate their queue in units of time not bytes. As Both AQMs regulate their queue in units of time rather than bytes.
already explained, this ensures configuration can be invariant for As already explained, this ensures configuration can be invariant for
different drain rates. With AQMs in a dualQ structure this is different drain rates. With AQMs in a dualQ structure this is
particularly important because the drain rate of each queue can vary particularly important because the drain rate of each queue can vary
rapidly as flows for the two queues arrive and depart, even if the rapidly as flows for the two queues arrive and depart, even if the
combined link rate is constant. combined link rate is constant.
It would be possible to control the queues with other alternative It would be possible to control the queues with other alternative
AQMs, as long as the normative requirements (those expressed in AQMs, as long as the normative requirements (those expressed in
capitals) in Section 2.5 are observed. capitals) in Section 2.5 are observed.
2.5. Normative Requirements for a DualQ Coupled AQM 2.5. Normative Requirements for a DualQ Coupled AQM
skipping to change at page 12, line 42 skipping to change at page 14, line 36
The following requirements are intended to capture only the essential The following requirements are intended to capture only the essential
aspects of a DualQ Coupled AQM. They are intended to be independent aspects of a DualQ Coupled AQM. They are intended to be independent
of the particular AQMs used for each queue. of the particular AQMs used for each queue.
2.5.1. Functional Requirements 2.5.1. Functional Requirements
A Dual Queue Coupled AQM implementation MUST utilize two queues, each A Dual Queue Coupled AQM implementation MUST utilize two queues, each
with an AQM algorithm. The two queues can be part of a larger with an AQM algorithm. The two queues can be part of a larger
queuing hierarchy [I-D.briscoe-tsvwg-l4s-diffserv]. queuing hierarchy [I-D.briscoe-tsvwg-l4s-diffserv].
The AQM algorithm for the low latency (L) queue MUST apply ECN The AQM algorithm for the low latency (L) queue MUST be able to apply
marking. ECN marking to ECN-capable packets.
The scheduler draining the two queues MUST give L4S packets priority The scheduler draining the two queues MUST give L4S packets priority
over Classic, although priority MUST be bounded in order not to over Classic, although priority MUST be bounded in order not to
starve Classic traffic. starve Classic traffic.
[I-D.ietf-tsvwg-ecn-l4s-id] defines the meaning of an ECN marking on [I-D.ietf-tsvwg-ecn-l4s-id] defines the meaning of an ECN marking on
L4S traffic, relative to drop of Classic traffic. In order to L4S traffic, relative to drop of Classic traffic. In order to ensure
prevent starvation of Classic traffic by scalable L4S traffic, it coexistence of Classic and Scalable L4S traffic, it says, "The
says, "The likelihood that an AQM drops a Not-ECT Classic packet likelihood that an AQM drops a Not-ECT Classic packet (p_C) MUST be
(p_C) MUST be roughly proportional to the square of the likelihood roughly proportional to the square of the likelihood that it would
that it would have marked it if it had been an L4S packet (p_L)." have marked it if it had been an L4S packet (p_L)." The term
The term 'likelihood' is used to allow for marking and dropping to be 'likelihood' is used to allow for marking and dropping to be either
either probabilistic or deterministic. probabilistic or deterministic.
For the current specification, this translates into the following For the current specification, this translates into the following
requirement. A DualQ Coupled AQM MUST apply ECN marking to traffic requirement. A DualQ Coupled AQM MUST apply ECN marking to traffic
in the L queue that is no lower than that derived from the likelihood in the L queue that is no lower than that derived from the likelihood
of drop (or ECN marking) in the Classic queue using Eqn. (1). of drop (or ECN marking) in the Classic queue using Eqn. (1).
The constant of proportionality, k, in Eqn (1) determines the The constant of proportionality, k, in Eqn (1) determines the
relative flow rates of Classic and L4S flows when the AQM concerned relative flow rates of Classic and L4S flows when the AQM concerned
is the bottleneck (all other factors being equal). is the bottleneck (all other factors being equal).
[I-D.ietf-tsvwg-ecn-l4s-id] says, "The constant of proportionality [I-D.ietf-tsvwg-ecn-l4s-id] says, "The constant of proportionality
(k) does not have to be standardised for interoperability, but a (k) does not have to be standardised for interoperability, but a
value of 2 is RECOMMENDED." value of 2 is RECOMMENDED."
Assuming scalable congestion controls for the Internet will be as Assuming Scalable congestion controls for the Internet will be as
aggressive as DCTCP, this will ensure their congestion window will be aggressive as DCTCP, this will ensure their congestion window will be
roughly the same as that of a standards track TCP congestion control roughly the same as that of a standards track TCP congestion control
(Reno) [RFC5681] and other so-called TCP-friendly controls, such as (Reno) [RFC5681] and other so-called TCP-friendly controls, such as
TCP Cubic in its TCP-friendly mode. TCP Cubic in its TCP-friendly mode.
The choice of k is a matter of operator policy, and operators MAY The choice of k is a matter of operator policy, and operators MAY
choose a different value using Table 1 and the guidelines in choose a different value using Table 1 and the guidelines in
Appendix C. Appendix C.
If multiple users share capacity at a bottleneck (e.g. in the If multiple customers or users share capacity at a bottleneck (e.g.
Internet access link of a campus network), the operator's choice of k in the Internet access link of a campus network), the operator's
will determine capacity sharing between the flows of different users. choice of k will determine capacity sharing between the flows of
However, on the public Internet, access network operators typically different customers. However, on the public Internet, access network
isolate customers from each other with some form of layer-2 operators typically isolate customers from each other with some form
multiplexing (OFDM(A) in DOCSIS3.1, CDMA in 3G, SC-FDMA in LTE) or L3 of layer-2 multiplexing (OFDM(A) in DOCSIS3.1, CDMA in 3G, SC-FDMA in
scheduling (WRR in DSL), rather than relying on TCP to share capacity LTE) or L3 scheduling (WRR in DSL), rather than relying on TCP to
between customers [RFC0970]. In such cases, the choice of k will share capacity between customers [RFC0970]. In such cases, the
solely affect relative flow rates within each customer's access choice of k will solely affect relative flow rates within each
capacity, not between customers. Also, k will not affect relative customer's access capacity, not between customers. Also, k will not
flow rates at any times when all flows are Classic or all L4S, and it affect relative flow rates at any times when all flows are Classic or
will not affect the relative throughput of small flows. all flows are L4S, and it will not affect the relative throughput of
small flows.
2.5.1.1. Requirements in Unexpected Cases 2.5.1.1. Requirements in Unexpected Cases
The flexibility to allow operator-specific classifiers (Section 2.3) The flexibility to allow operator-specific classifiers (Section 2.3)
leads to the need to specify what the AQM in each queue ought to do leads to the need to specify what the AQM in each queue ought to do
with packets that do not carry the ECN field expected for that queue. with packets that do not carry the ECN field expected for that queue.
It is recommended that the AQM in each queue inspects the ECN field It is recommended that the AQM in each queue inspects the ECN field
to determine what sort of congestion notification to signal, then to determine what sort of congestion notification to signal, then
decides whether to apply congestion notification to this particular decides whether to apply congestion notification to this particular
packet, as follows: packet, as follows:
skipping to change at page 14, line 38 skipping to change at page 16, line 33
SHOULD apply drop using a drop probability appropriate to SHOULD apply drop using a drop probability appropriate to
Classic congestion control and appropriate to the target Classic congestion control and appropriate to the target
delay in the L queue delay in the L queue
o If a packet that carries an ECT(1) codepoint is classified into o If a packet that carries an ECT(1) codepoint is classified into
the C queue: the C queue:
* the C AQM SHOULD apply CE-marking using the coupled AQM * the C AQM SHOULD apply CE-marking using the coupled AQM
probability p_CL (= k*p'). probability p_CL (= k*p').
If the DualQ Coupled AQM has detected overload, it will signal
congestion solely using drop, irrespective of the ECN field.
The above requirements are worded as "SHOULDs", because operator- The above requirements are worded as "SHOULDs", because operator-
specific classifiers are for flexibility, by definition. Therefore, specific classifiers are for flexibility, by definition. Therefore,
alternative actions might be appropriate in the operator's specific alternative actions might be appropriate in the operator's specific
circumstances. An example would be where the operator knows that circumstances. An example would be where the operator knows that
certain legacy traffic marked with one codepoint actually has a certain legacy traffic marked with one codepoint actually has a
congestion response associated with another codepoint. congestion response associated with another codepoint.
If the DualQ Coupled AQM has detected overload, it MUST signal
congestion solely using drop, irrespective of the ECN field.
Switching to drop if ECN marking is persistently high is required by
Section 7 of [RFC3168] and Section 4.2.1 of [RFC7567].
2.5.2. Management Requirements 2.5.2. Management Requirements
2.5.2.1. Configuration
By default, a DualQ Coupled AQM SHOULD NOT need any configuration for By default, a DualQ Coupled AQM SHOULD NOT need any configuration for
use at a bottleneck on the public Internet [RFC7567]. The following use at a bottleneck on the public Internet [RFC7567]. The following
parameters MAY be operator-configurable, e.g. to tune for non- parameters MAY be operator-configurable, e.g. to tune for non-
Internet settings: Internet settings:
o Optional packet classifier(s) to use in addition to the ECN field o Optional packet classifier(s) to use in addition to the ECN field
(see Section 2.3); (see Section 2.3);
o Expected typical RTT (a parameter for typical or target queuing o Expected typical RTT, which can be used to determine the queuing
delay in each queue might be configurable instead; if so it MUST delay of the Classic AQM at its operating point, in order to
be expressed in units of time); prevent typical lone TCP flows from under-utilizing capacity. For
example:
o Expected maximum RTT (a stability parameter that depends on * for the PI2 algorithm (Appendix A) the queuing delay target is
maximum RTT might be configurable instead); set to the typical RTT;
* for the Curvy RED algorithm (Appendix B) the queuing delay at
the desired operating point of the curvy ramp is configured to
encompass a typical RTT;
* if another Classic AQM was used, it would be likely to need an
operating point for the queue based on the typical RTT, and if
so it SHOULD be expressed in units of time.
An operating point that is manually calculated might be directly
configurable instead, e.g. for links with large numbers of flows
where under-utilization by a single TCP flow would be unlikely.
o Expected maximum RTT, which can be used to set the stability
parameter(s) of the Classic AQM. For example:
* for the PI2 algorithm (Appendix A), the gain parameters of the
PI algorithm depend on the maximum RTT.
* for the Curvy RED algorithm (Appendix B) the smoothing
parameter is chosen to filter out transients in the queue
within a maximum RTT.
Stability parameter(s) that are manually calculated assuming a
maximum RTT might be directly configurable instead.
o Coupling factor, k; o Coupling factor, k;
o The limit to the conditional priority of L4S (scheduler-dependent, o A limit to the conditional priority of L4S. This is scheduler-
e.g. the scheduler weight for WRR, or the time-shift for time- dependent, but it SHOULD be expressed as a relation between the
shifted FIFO); max delay of a C packet and an L packet. For example:
* for a WRR scheduler a weight ratio between L and C of w:1 means
that the maximum delay to a C packet is w times that of an L
packet.
* for a time-shifted FIFO (TS-FIFO) scheduler (see Section 4.1.1)
a time-shift of tshift means that the maximum delay to a C
packet is tshift greater than that of an L packet. tshift could
be expressed as a multiple of the typical RTT rather than as an
absolute delay.
o The maximum Classic ECN marking probability, p_Cmax, before o The maximum Classic ECN marking probability, p_Cmax, before
switching over to drop. switching over to drop.
2.5.2.2. Monitoring
An experimental DualQ Coupled AQM SHOULD allow the operator to An experimental DualQ Coupled AQM SHOULD allow the operator to
monitor each of the following operational statistics on demand, per monitor each of the following operational statistics on demand, per
queue and per configurable sample interval, for performance queue and per configurable sample interval, for performance
monitoring and perhaps also for accounting in some cases: monitoring and perhaps also for accounting in some cases:
o Bits forwarded, from which utilization can be calculated; o Bits forwarded, from which utilization can be calculated;
o Total packets arriving, enqueued and dequeued to distinguish tail o Total packets in the three categories: arrived, presented to the
discard from proactive AQM discard; AQM, and forwarded. The difference between the first two will
measure any non-AQM tail discard. The difference between the last
two will measure proactive AQM discard;
o ECN packets marked, non-ECN packets dropped, ECN packets dropped, o ECN packets marked, non-ECN packets dropped, ECN packets dropped,
from which marking and dropping probabilities can be calculated; which can be combined with the three total packet counts above to
calculate marking and dropping probabilities;
o Queue delay (not including serialization delay of the head packet o Queue delay (not including serialization delay of the head packet
or medium acquisition delay) - see further notes below. or medium acquisition delay) - see further notes below.
Unlike the other statistics, queue delay cannot be captured in a Unlike the other statistics, queue delay cannot be captured in a
simple accumulating counter. Therefore the type of queue delay simple accumulating counter. Therefore the type of queue delay
statistics produced (mean, percentiles, etc.) will depend on statistics produced (mean, percentiles, etc.) will depend on
implementation constraints. To facilitate comparative evaluation implementation constraints. To facilitate comparative evaluation
of different implementations and approaches, an implementation of different implementations and approaches, an implementation
SHOULD allow mean and 99th percentile queue delay to be derived SHOULD allow mean and 99th percentile queue delay to be derived
(per queue per sample interval). A relatively simple way to do (per queue per sample interval). A relatively simple way to do
this would be to store a coarse-grained histogram of queue delay. this would be to store a coarse-grained histogram of queue delay.
This could be done with a small number of bins with configurable This could be done with a small number of bins with configurable
edges that represent contiguous ranges of queue delay. Then, over edges that represent contiguous ranges of queue delay. Then, over
a sample interval, each bin would accumulate a count of the number a sample interval, each bin would accumulate a count of the number
of packets that had fallen within each range. The maximum queue of packets that had fallen within each range. The maximum queue
delay per queue per interval MAY also be recorded. delay per queue per interval MAY also be recorded.
2.5.2.3. Anomaly Detection
An experimental DualQ Coupled AQM SHOULD asynchronously report the An experimental DualQ Coupled AQM SHOULD asynchronously report the
following data about anomalous conditions: following data about anomalous conditions:
o Start-time and duration of overload state. o Start-time and duration of overload state.
A hysteresis mechanism SHOULD be used to prevent flapping in and A hysteresis mechanism SHOULD be used to prevent flapping in and
out of overload causing an event storm. For instance, exit from out of overload causing an event storm. For instance, exit from
overload state could trigger one report, but also latch a timer. overload state could trigger one report, but also latch a timer.
Then, during that time, if the AQM enters and exits overload state Then, during that time, if the AQM enters and exits overload state
any number of times, the duration in overload state is accumulated any number of times, the duration in overload state is accumulated
but no new report is generated until the first time the AQM is out but no new report is generated until the first time the AQM is out
of overload once the timer has expired. of overload once the timer has expired.
2.5.2.4. Deployment, Coexistence and Scaling
[RFC5706] suggests that deployment, coexistence and scaling should [RFC5706] suggests that deployment, coexistence and scaling should
also be covered as management requirements. The raison d'etre of the also be covered as management requirements. The raison d'etre of the
DualQ Couple AQM is to enable deployment and coexistence of scalable DualQ Coupled AQM is to enable deployment and coexistence of Scalable
congestion controls - as incremental replacements for today's TCP- congestion controls - as incremental replacements for today's TCP-
friendly controls that do not scale with bandwidth-delay product. friendly controls that do not scale with bandwidth-delay product.
Therefore, these motivating issues are explained in the Introduction Therefore there is no need to repeat these motivating issues here
and detailed in the L4S architecture [I-D.ietf-tsvwg-l4s-arch]. given they are already explained in the Introduction and detailed in
Also, the descriptions of specific DualQ Coupled AQM algorithms in the L4S architecture [I-D.ietf-tsvwg-l4s-arch].
the appendices cover scaling of their configuration parameters, e.g.
with respect to RTT and sampling frequency. The descriptions of specific DualQ Coupled AQM algorithms in the
appendices cover scaling of their configuration parameters, e.g. with
respect to RTT and sampling frequency.
3. IANA Considerations 3. IANA Considerations
This specification contains no IANA considerations. This specification contains no IANA considerations.
4. Security Considerations 4. Security Considerations
4.1. Overload Handling 4.1. Overload Handling
Where the interests of users or flows might conflict, it could be Where the interests of users or flows might conflict, it could be
skipping to change at page 17, line 22 skipping to change at page 20, line 11
Under overload the higher priority L4S service will have to sacrifice Under overload the higher priority L4S service will have to sacrifice
some aspect of its performance. Alternative solutions are provided some aspect of its performance. Alternative solutions are provided
below that each relax a different factor: e.g. throughput, delay, below that each relax a different factor: e.g. throughput, delay,
drop. These choices need to be made either by the developer or by drop. These choices need to be made either by the developer or by
operator policy, rather than by the IETF. operator policy, rather than by the IETF.
4.1.1. Avoiding Classic Starvation: Sacrifice L4S Throughput or Delay? 4.1.1. Avoiding Classic Starvation: Sacrifice L4S Throughput or Delay?
Priority of L4S is required to be conditional to avoid total Priority of L4S is required to be conditional to avoid total
throughput starvation of Classic by heavy L4S traffic. This raises starvation of Classic by heavy L4S traffic. This raises the question
the question of whether to sacrifice L4S throughput or L4S delay (or of whether to sacrifice L4S throughput or L4S delay (or some other
some other policy) to mitigate starvation of Classic: policy) to mitigate starvation of Classic:
Sacrifice L4S throughput: By using weighted round robin as the Sacrifice L4S throughput: By using weighted round robin as the
conditional priority scheduler, the L4S service can sacrifice some conditional priority scheduler, the L4S service can sacrifice some
throughput during overload to guarantee a minimum throughput throughput during overload. This can either be thought of as
service for Classic traffic. The scheduling weight of the Classic guaranteeing a minimum throughput service for Classic traffic, or
queue should be small (e.g. 1/16). Then, in most traffic as guaranteeing a maximum delay for a packet at the head of the
scenarios the scheduler will not interfere and it will not need to Classic queue.
- the coupling mechanism and the end-systems will share out the
capacity across both queues as if it were a single pool. However, The scheduling weight of the Classic queue should be small (e.g.
because the congestion coupling only applies in one direction 1/16). Then, in most traffic scenarios the scheduler will not
(from C to L), if L4S traffic is over-aggressive or unresponsive, interfere and it will not need to - the coupling mechanism and the
the scheduler weight for Classic traffic will at least be large end-systems will share out the capacity across both queues as if
enough to ensure it does not starve. it were a single pool. However, because the congestion coupling
only applies in one direction (from C to L), if L4S traffic is
over-aggressive or unresponsive, the scheduler weight for Classic
traffic will at least be large enough to ensure it does not
starve.
In cases where the ratio of L4S to Classic flows (e.g. 19:1) is In cases where the ratio of L4S to Classic flows (e.g. 19:1) is
greater than the ratio of their scheduler weights (e.g. 15:1), the greater than the ratio of their scheduler weights (e.g. 15:1), the
L4S flows will get less than an equal share of the capacity, but L4S flows will get less than an equal share of the capacity, but
only slightly. For instance, with the example numbers given, each only slightly. For instance, with the example numbers given, each
L4S flow will get (15/16)/19 = 4.9% when ideally each would get L4S flow will get (15/16)/19 = 4.9% when ideally each would get
1/20=5%. In the rather specific case of an unresponsive flow 1/20=5%. In the rather specific case of an unresponsive flow
taking up a large part of the capacity set aside for L4S, using taking up just less than the capacity set aside for L4S (e.g.
WRR could significantly reduce the capacity left for any 14/16 in the above example), using WRR could significantly reduce
responsive L4S flows. the capacity left for any responsive L4S flows.
The scheduling weight of the Classic queue should not be too
small, otherwise a C packet at the head of the queue could be
excessively delayed by a continually busy L queue. For instance
if the Classic weight is 1/16, the maximum that a Classic packet
at the head of the queue can be delayed by L traffic is the
serialization delay of 15 MTU-sized packets.
Sacrifice L4S Delay: To control milder overload of responsive Sacrifice L4S Delay: To control milder overload of responsive
traffic, particularly when close to the maximum congestion signal, traffic, particularly when close to the maximum congestion signal,
the operator could choose to control overload of the Classic queue the operator could choose to control overload of the Classic queue
by allowing some delay to 'leak' across to the L4S queue. The by allowing some delay to 'leak' across to the L4S queue. The
scheduler can be made to behave like a single First-In First-Out scheduler can be made to behave like a single First-In First-Out
(FIFO) queue with different service times by implementing a very (FIFO) queue with different service times by implementing a very
simple conditional priority scheduler that could be called a simple conditional priority scheduler that could be called a
"time-shifted FIFO" (see the Modifier Earliest Deadline First "time-shifted FIFO" (see the Modifier Earliest Deadline First
(MEDF) scheduler of [MEDF]). This scheduler adds tshift to the (MEDF) scheduler of [MEDF]). This scheduler adds tshift to the
queue delay of the next L4S packet, before comparing it with the queue delay of the next L4S packet, before comparing it with the
queue delay of the next Classic packet, then it selects the packet queue delay of the next Classic packet, then it selects the packet
with the greater adjusted queue delay. Under regular conditions, with the greater adjusted queue delay. Under regular conditions,
this time-shifted FIFO scheduler behaves just like a strict this time-shifted FIFO scheduler behaves just like a strict
priority scheduler. But under moderate or high overload it priority scheduler. But under moderate or high overload it
prevents starvation of the Classic queue, because the time-shift prevents starvation of the Classic queue, because the time-shift
(tshift) defines the maximum extra queuing delay of Classic (tshift) defines the maximum extra queuing delay of Classic
packets relative to L4S. packets relative to L4S.
The example implementation in Appendix A can implement either policy. The example implementations in Appendix A and Appendix B could both
be implemented with either policy.
4.1.2. Congestion Signal Saturation: Introduce L4S Drop or Delay? 4.1.2. Congestion Signal Saturation: Introduce L4S Drop or Delay?
To keep the throughput of both L4S and Classic flows roughly equal To keep the throughput of both L4S and Classic flows roughly equal
over the full load range, a different control strategy needs to be over the full load range, a different control strategy needs to be
defined above the point where one AQM first saturates to a defined above the point where one AQM first saturates to a
probability of 100% leaving no room to push back the load any harder. probability of 100% leaving no room to push back the load any harder.
If k>1, L4S will saturate first, even though saturation could be If k>1, L4S will saturate first, even though saturation could be
caused by unresponsive traffic in either queue. caused by unresponsive traffic in either queue.
skipping to change at page 19, line 38 skipping to change at page 22, line 38
Experiments with the DualPI2 AQM (Appendix A) have shown that Experiments with the DualPI2 AQM (Appendix A) have shown that
introducing 'drop on saturation' at 100% L4S marking addresses this introducing 'drop on saturation' at 100% L4S marking addresses this
problem with unresponsive ECN as well as addressing the saturation problem with unresponsive ECN as well as addressing the saturation
problem. It leaves only a small range of congestion levels where problem. It leaves only a small range of congestion levels where
unresponsive traffic gains any advantage from using the ECN unresponsive traffic gains any advantage from using the ECN
capability, and the advantage is hardly detectable [DualQ-Test]. capability, and the advantage is hardly detectable [DualQ-Test].
5. Acknowledgements 5. Acknowledgements
Thanks to Anil Agarwal, Sowmini Varadhan's and Gabi Bracha for Thanks to Anil Agarwal, Sowmini Varadhan's, Gabi Bracha, Nicolas
detailed review comments particularly of the appendices and Kuhn, Tom Henderson and David Pullen for detailed review comments
suggestions on how to make our explanation clearer. Thanks also to particularly of the appendices and suggestions on how to make the
Greg White for improving the normative requirements and both Greg and explanations clearer. Thanks also to Tom Henderson for insights on
Tom Henderson for insights on the choice of schedulers, queue delay the choice of schedulers and queue delay measurement techniques.
measurement techniques.
The authors' contributions were originally part-funded by the The early contributions of Koen De Schepper, Bob Briscoe, Olga
European Community under its Seventh Framework Programme through the Bondarenko and Inton Tsang were part-funded by the European Community
Reducing Internet Transport Latency (RITE) project (ICT-317700). Bob under its Seventh Framework Programme through the Reducing Internet
Briscoe's contribution was also part-funded by the Research Council Transport Latency (RITE) project (ICT-317700). Bob Briscoe's
of Norway through the TimeIn project. The views expressed here are contribution was also part-funded by the Research Council of Norway
solely those of the authors. through the TimeIn project. The views expressed here are solely
those of the authors.
6. References 6. Contributors
6.1. Normative References The following contributed implementations and evaluations that
validated and helped to improve this specification:
Olga Albisser <olga@albisser.org> of Simula Research Lab, Norway
(Olga Bondarenko during early drafts) implemented the prototype
DualPI2 AQM for Linux with Koen De Schepper and conducted
extensive evaluations as well as implementing the live performance
visualization GUI [L4Sdemo16].
Olivier Tilmans <olivier.tilmans@nokia-bell-labs.com> of Nokia
Bell Labs, Belgium prepared and maintains the Linux implementation
of DualPI2 for upstreaming.
Tom Henderson <tomh@tomh.org> of the University of Washington, WA,
US implemented various Coupled DualQ AQMs for ns3, including
DualPI2 and DualPIE over point to point and DOCSIS 3.1 link models
and conducted extensive evaluations.
Ing Jyh (Inton) Tsang of Nokia, Belgium built the End-to-End Data
Centre to the Home broadband testbed on which Coupled DualQ
implementations were tested.
7. References
7.1. Normative References
[I-D.ietf-tsvwg-ecn-l4s-id]
Schepper, K. and B. Briscoe, "Identifying Modified
Explicit Congestion Notification (ECN) Semantics for
Ultra-Low Queuing Delay (L4S)", draft-ietf-tsvwg-ecn-l4s-
id-06 (work in progress), March 2019.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997, DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>. <https://www.rfc-editor.org/info/rfc2119>.
6.2. Informative References [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition
of Explicit Congestion Notification (ECN) to IP",
RFC 3168, DOI 10.17487/RFC3168, September 2001,
<https://www.rfc-editor.org/info/rfc3168>.
[RFC8311] Black, D., "Relaxing Restrictions on Explicit Congestion
Notification (ECN) Experimentation", RFC 8311,
DOI 10.17487/RFC8311, January 2018,
<https://www.rfc-editor.org/info/rfc8311>.
7.2. Informative References
[Alizadeh-stability]
Alizadeh, M., Javanmard, A., and B. Prabhakar, "Analysis
of DCTCP: Stability, Convergence, and Fairness", ACM
SIGMETRICS 2011 , June 2011,
<https://dl.acm.org/citation.cfm?id=1993753>.
[AQMmetrics]
Kwon, M. and S. Fahmy, "A Comparison of Load-based and
Queue- based Active Queue Management Algorithms", Proc.
Int'l Soc. for Optical Engineering (SPIE) 4866:35--46 DOI:
10.1117/12.473021, 2002,
<https://www.cs.purdue.edu/homes/fahmy/papers/ldc.pdf>.
[ARED01] Floyd, S., Gummadi, R., and S. Shenker, "Adaptive RED: An [ARED01] Floyd, S., Gummadi, R., and S. Shenker, "Adaptive RED: An
Algorithm for Increasing the Robustness of RED's Active Algorithm for Increasing the Robustness of RED's Active
Queue Management", ACIRI Technical Report , August 2001, Queue Management", ACIRI Technical Report , August 2001,
<http://www.icir.org/floyd/red.html>. <http://www.icir.org/floyd/red.html>.
[BBRv1] Cardwell, N., Cheng, Y., Hassas Yeganeh, S., and V.
Jacobson, "BBR Congestion Control", Internet Draft draft-
cardwell-iccrg-bbr-congestion-control-00, July 2017,
<https://tools.ietf.org/html/
draft-cardwell-iccrg-bbr-congestion-control-00>.
[CoDel] Nichols, K. and V. Jacobson, "Controlling Queue Delay", [CoDel] Nichols, K. and V. Jacobson, "Controlling Queue Delay",
ACM Queue 10(5), May 2012, ACM Queue 10(5), May 2012,
<http://queue.acm.org/issuedetail.cfm?issue=2208917>. <http://queue.acm.org/issuedetail.cfm?issue=2208917>.
[CRED_Insights] [CRED_Insights]
Briscoe, B., "Insights from Curvy RED (Random Early Briscoe, B., "Insights from Curvy RED (Random Early
Detection)", BT Technical Report TR-TUB8-2015-003, July Detection)", BT Technical Report TR-TUB8-2015-003
2015, arXiv:1904.07339 [cs.NI], July 2015,
<http://www.bobbriscoe.net/projects/latency/credi_tr.pdf>. <https://arxiv.org/abs/1904.07339>.
[DCttH15] De Schepper, K., Bondarenko, O., Briscoe, B., and I. [DCttH15] De Schepper, K., Bondarenko, O., Briscoe, B., and I.
Tsang, "`Data Centre to the Home': Ultra-Low Latency for Tsang, "`Data Centre to the Home': Ultra-Low Latency for
All", 2015, <http://www.bobbriscoe.net/projects/latency/ All", RITE project Technical Report , 2015,
dctth_preprint.pdf>. <http://riteproject.eu/publications/>.
(Under submission) [DOCSIS3.1]
CableLabs, "MAC and Upper Layer Protocols Interface
(MULPI) Specification, CM-SP-MULPIv3.1", Data-Over-Cable
Service Interface Specifications DOCSIS(R) 3.1 Version i17
or later, January 2019, <https://specification-
search.cablelabs.com/CM-SP-MULPIv3.1>.
[DualPI2Linux]
Albisser, O., De Schepper, K., Briscoe, B., Tilmans, O.,
and H. Steen, "DUALPI2 - Low Latency, Low Loss and
Scalable (L4S) AQM", Proc. Linux Netdev 0x13 , March 2019,
<https://www.netdevconf.org/0x13/
session.html?talk-DUALPI2-AQM>.
[DualQ-Test] [DualQ-Test]
Steen, H., "Destruction Testing: Ultra-Low Delay using Steen, H., "Destruction Testing: Ultra-Low Delay using
Dual Queue Coupled Active Queue Management", Masters Dual Queue Coupled Active Queue Management", Masters
Thesis, Dept of Informatics, Uni Oslo , May 2017. Thesis, Dept of Informatics, Uni Oslo , May 2017.
[I-D.briscoe-tsvwg-l4s-diffserv] [I-D.briscoe-tsvwg-l4s-diffserv]
Briscoe, B., "Interactions between Low Latency, Low Loss, Briscoe, B., "Interactions between Low Latency, Low Loss,
Scalable Throughput (L4S) and Differentiated Services", Scalable Throughput (L4S) and Differentiated Services",
draft-briscoe-tsvwg-l4s-diffserv-00 (work in progress), draft-briscoe-tsvwg-l4s-diffserv-02 (work in progress),
March 2018. November 2018.
[I-D.ietf-tsvwg-ecn-l4s-id]
Schepper, K., Briscoe, B., and I. Tsang, "Identifying
Modified Explicit Congestion Notification (ECN) Semantics
for Ultra-Low Queuing Delay", draft-ietf-tsvwg-ecn-l4s-
id-02 (work in progress), March 2018.
[I-D.ietf-tsvwg-l4s-arch] [I-D.ietf-tsvwg-l4s-arch]
Briscoe, B., Schepper, K., and M. Bagnulo, "Low Latency, Briscoe, B., Schepper, K., and M. Bagnulo, "Low Latency,
Low Loss, Scalable Throughput (L4S) Internet Service: Low Loss, Scalable Throughput (L4S) Internet Service:
Architecture", draft-ietf-tsvwg-l4s-arch-02 (work in Architecture", draft-ietf-tsvwg-l4s-arch-03 (work in
progress), March 2018. progress), October 2018.
[I-D.sridharan-tcpm-ctcp]
Sridharan, M., Tan, K., Bansal, D., and D. Thaler,
"Compound TCP: A New TCP Congestion Control for High-Speed
and Long Distance Networks", draft-sridharan-tcpm-ctcp-02
(work in progress), November 2008.
[L4Sdemo16] [L4Sdemo16]
Bondarenko, O., De Schepper, K., Tsang, I., and B. Bondarenko, O., De Schepper, K., Tsang, I., and B.
Briscoe, "Ultra-Low Delay for All: Live Experience, Live Briscoe, "Ultra-Low Delay for All: Live Experience, Live
Analysis", Proc. MMSYS'16 pp33:1--33:4, May 2016, Analysis", Proc. MMSYS'16 pp33:1--33:4, May 2016,
<http://dl.acm.org/citation.cfm?doid=2910017.2910633 <http://dl.acm.org/citation.cfm?doid=2910017.2910633
(videos of demos: https://riteproject.eu/ (videos of demos: https://riteproject.eu/
dctth/#1511dispatchwg )>. dctth/#1511dispatchwg )>.
[LLD] White, G., Sundaresan, K., and B. Briscoe, "Low Latency
DOCSIS: Technology Overview", CableLabs White Paper ,
February 2019, <https://cablela.bs/
low-latency-docsis-technology-overview-february-2019>.
[Mathis09] [Mathis09]
Mathis, M., "Relentless Congestion Control", PFLDNeT'09 , Mathis, M., "Relentless Congestion Control", PFLDNeT'09 ,
May 2009, <http://www.hpcc.jp/pfldnet2009/ May 2009, <http://www.hpcc.jp/pfldnet2009/
Program_files/1569198525.pdf>. Program_files/1569198525.pdf>.
[MEDF] Menth, M., Schmid, M., Heiss, H., and T. Reim, "MEDF - a [MEDF] Menth, M., Schmid, M., Heiss, H., and T. Reim, "MEDF - a
simple scheduling algorithm for two real-time transport simple scheduling algorithm for two real-time transport
service classes with application in the UTRAN", Proc. IEEE service classes with application in the UTRAN", Proc. IEEE
Conference on Computer Communications (INFOCOM'03) Vol.2 Conference on Computer Communications (INFOCOM'03) Vol.2
pp.1116-1122, March 2003. pp.1116-1122, March 2003.
[PI2] De Schepper, K., Bondarenko, O., Briscoe, B., and I. [PI2] De Schepper, K., Bondarenko, O., Briscoe, B., and I.
Tsang, "PI2: A Linearized AQM for both Classic and Tsang, "PI2: A Linearized AQM for both Classic and
Scalable TCP", ACM CoNEXT'16 , December 2016, Scalable TCP", ACM CoNEXT'16 , December 2016,
<https://riteproject.files.wordpress.com/2015/10/ <https://riteproject.files.wordpress.com/2015/10/
pi2_conext.pdf>. pi2_conext.pdf>.
(To appear) [PragueLinux]
Briscoe, B., De Schepper, K., Albisser, O., Misund, J.,
Tilmans, O., Kuehlewind, M., and A. Ahmed, "Implementing
the `TCP Prague' Requirements for Low Latency Low Loss
Scalable Throughput (L4S)", Proc. Linux Netdev 0x13 ,
March 2019, <https://www.netdevconf.org/0x13/
session.html?talk-tcp-prague-l4s>.
[RFC0970] Nagle, J., "On Packet Switches With Infinite Storage", [RFC0970] Nagle, J., "On Packet Switches With Infinite Storage",
RFC 970, DOI 10.17487/RFC0970, December 1985, RFC 970, DOI 10.17487/RFC0970, December 1985,
<https://www.rfc-editor.org/info/rfc970>. <https://www.rfc-editor.org/info/rfc970>.
[RFC2309] Braden, B., Clark, D., Crowcroft, J., Davie, B., Deering, [RFC2309] Braden, B., Clark, D., Crowcroft, J., Davie, B., Deering,
S., Estrin, D., Floyd, S., Jacobson, V., Minshall, G., S., Estrin, D., Floyd, S., Jacobson, V., Minshall, G.,
Partridge, C., Peterson, L., Ramakrishnan, K., Shenker, Partridge, C., Peterson, L., Ramakrishnan, K., Shenker,
S., Wroclawski, J., and L. Zhang, "Recommendations on S., Wroclawski, J., and L. Zhang, "Recommendations on
Queue Management and Congestion Avoidance in the Queue Management and Congestion Avoidance in the
Internet", RFC 2309, DOI 10.17487/RFC2309, April 1998, Internet", RFC 2309, DOI 10.17487/RFC2309, April 1998,
<https://www.rfc-editor.org/info/rfc2309>. <https://www.rfc-editor.org/info/rfc2309>.
[RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition
of Explicit Congestion Notification (ECN) to IP",
RFC 3168, DOI 10.17487/RFC3168, September 2001,
<https://www.rfc-editor.org/info/rfc3168>.
[RFC3246] Davie, B., Charny, A., Bennet, J., Benson, K., Le Boudec, [RFC3246] Davie, B., Charny, A., Bennet, J., Benson, K., Le Boudec,
J., Courtney, W., Davari, S., Firoiu, V., and D. J., Courtney, W., Davari, S., Firoiu, V., and D.
Stiliadis, "An Expedited Forwarding PHB (Per-Hop Stiliadis, "An Expedited Forwarding PHB (Per-Hop
Behavior)", RFC 3246, DOI 10.17487/RFC3246, March 2002, Behavior)", RFC 3246, DOI 10.17487/RFC3246, March 2002,
<https://www.rfc-editor.org/info/rfc3246>. <https://www.rfc-editor.org/info/rfc3246>.
[RFC3649] Floyd, S., "HighSpeed TCP for Large Congestion Windows", [RFC3649] Floyd, S., "HighSpeed TCP for Large Congestion Windows",
RFC 3649, DOI 10.17487/RFC3649, December 2003, RFC 3649, DOI 10.17487/RFC3649, December 2003,
<https://www.rfc-editor.org/info/rfc3649>. <https://www.rfc-editor.org/info/rfc3649>.
skipping to change at page 23, line 16 skipping to change at page 27, line 33
and G. Judd, "Data Center TCP (DCTCP): TCP Congestion and G. Judd, "Data Center TCP (DCTCP): TCP Congestion
Control for Data Centers", RFC 8257, DOI 10.17487/RFC8257, Control for Data Centers", RFC 8257, DOI 10.17487/RFC8257,
October 2017, <https://www.rfc-editor.org/info/rfc8257>. October 2017, <https://www.rfc-editor.org/info/rfc8257>.
[RFC8290] Hoeiland-Joergensen, T., McKenney, P., Taht, D., Gettys, [RFC8290] Hoeiland-Joergensen, T., McKenney, P., Taht, D., Gettys,
J., and E. Dumazet, "The Flow Queue CoDel Packet Scheduler J., and E. Dumazet, "The Flow Queue CoDel Packet Scheduler
and Active Queue Management Algorithm", RFC 8290, and Active Queue Management Algorithm", RFC 8290,
DOI 10.17487/RFC8290, January 2018, DOI 10.17487/RFC8290, January 2018,
<https://www.rfc-editor.org/info/rfc8290>. <https://www.rfc-editor.org/info/rfc8290>.
[RFC8311] Black, D., "Relaxing Restrictions on Explicit Congestion [RFC8298] Johansson, I. and Z. Sarker, "Self-Clocked Rate Adaptation
Notification (ECN) Experimentation", RFC 8311, for Multimedia", RFC 8298, DOI 10.17487/RFC8298, December
DOI 10.17487/RFC8311, January 2018, 2017, <https://www.rfc-editor.org/info/rfc8298>.
<https://www.rfc-editor.org/info/rfc8311>.
[RFC8312] Rhee, I., Xu, L., Ha, S., Zimmermann, A., Eggert, L., and [RFC8312] Rhee, I., Xu, L., Ha, S., Zimmermann, A., Eggert, L., and
R. Scheffenegger, "CUBIC for Fast Long-Distance Networks", R. Scheffenegger, "CUBIC for Fast Long-Distance Networks",
RFC 8312, DOI 10.17487/RFC8312, February 2018, RFC 8312, DOI 10.17487/RFC8312, February 2018,
<https://www.rfc-editor.org/info/rfc8312>. <https://www.rfc-editor.org/info/rfc8312>.
[TCP-CA] Jacobson, V. and M. Karels, "Congestion Avoidance and [SigQ-Dyn]
Control", Laurence Berkeley Labs Technical Report , Briscoe, B., "Rapid Signalling of Queue Dynamics",
November 1988, <http://ee.lbl.gov/papers/congavoid.pdf>. Technical Report TR-BB-2017-001 arXiv:1904.07044 [cs.NI],
September 2017, <https://arxiv.org/abs/1904.07044>.
Appendix A. Example DualQ Coupled PI2 Algorithm Appendix A. Example DualQ Coupled PI2 Algorithm
As a first concrete example, the pseudocode below gives the DualPI2 As a first concrete example, the pseudocode below gives the DualPI2
algorithm. DualPI2 follows the structure of the DualQ Coupled AQM algorithm. DualPI2 follows the structure of the DualQ Coupled AQM
framework in Figure 1. A simple step threshold (in units of queuing framework in Figure 1. A simple ramp function (configured in units
time) is used for the Native L4S AQM, but a ramp is also described as of queuing time) with unsmoothed ECN marking is used for the Native
an alternative. And the PI2 algorithm [PI2] is used for the Classic L4S AQM. The ramp can also be configured as a step function. The
AQM. PI2 is an improved variant of the PIE AQM [RFC8033]. PI2 algorithm [PI2] is used for the Classic AQM. PI2 is an improved
variant of the PIE AQM [RFC8033].
We will introduce the pseudocode in two passes. The first pass The pseudocode will be introduced in two passes. The first pass
explains the core concepts, deferring handling of overload to the explains the core concepts, deferring handling of overload to the
second pass. To aid comparison, line numbers are kept in step second pass. To aid comparison, line numbers are kept in step
between the two passes by using letter suffixes where the longer code between the two passes by using letter suffixes where the longer code
needs extra lines. needs extra lines.
All variables are assumed to be floating point in their basic units
(size in bytes, time in seconds, rates in bytes/second, alpha and
beta in Hz, and probabilities from 0 to 1. Constants expressed in k,
M, G, u, m, %, ... are assumed to be converted to their appropriate
multiple or fraction. A real implementation that wants to use
integer values needs to handle appropriate scaling factors and allow
accordingly appropriate resolution of its integer types (including
temporary internal values during calculations).
A full open source implementation for Linux is available at: A full open source implementation for Linux is available at:
https://github.com/olgabo/dualpi2. https://github.com/L4STeam/sch_dualpi2_upstream and explained in
[DualPI2Linux]. The specification of the DualQ Coupled AQM for
DOCSIS cable modems and CMTSs is available in [DOCSIS3.1] and
explained in [LLD].
A.1. Pass #1: Core Concepts A.1. Pass #1: Core Concepts
The pseudocode manipulates three main structures of variables: the The pseudocode manipulates three main structures of variables: the
packet (pkt), the L4S queue (lq) and the Classic queue (cq). The packet (pkt), the L4S queue (lq) and the Classic queue (cq). The
pseudocode consists of the following five functions: pseudocode consists of the following six functions:
o initialization code (Figure 2) that sets parameter defaults (the o the initialization function dualpi2_params_init(...) (Figure 2)
API for setting non-default values is omitted for brevity) that sets parameter defaults (the API for setting non-default
values is omitted for brevity)
o enqueue code (Figure 3) o the enqueue function dualpi2_enqueue(lq, cq, pkt) (Figure 3)
o dequeue code (Figure 4) o the dequeue function dualpi2_dequeue(lq, cq, pkt) (Figure 4)
o a ramp function (Figure 5) used to calculate the ECN-marking o recur(likelihood) for de-randomized ECN marking (shown at the end
probability for the L4S queue of Figure 4).
o code to regularly update the base probability (p) used in the o the L4S AQM function laqm(qdelay) (Figure 5) used to calculate the
dequeue code (Figure 6). ECN-marking probability for the L4S queue
o the base AQM function that implements the PI algorithm
dualpi2_update(lq, cq) (Figure 6) used to regularly update the
base probability (p'), which is squared for the Classic AQM as
well as being coupled across to the L4S queue.
It also uses the following functions that are not shown in full here: It also uses the following functions that are not shown in full here:
o scheduler(), which selects between the head packets of the two o scheduler(), which selects between the head packets of the two
queues; the choice of scheduler technology is discussed later; queues; the choice of scheduler technology is discussed later;
o cq.len() or lq.len() returns the current length (aka. backlog) of o cq.len() or lq.len() returns the current length (aka. backlog) of
the relevant queue in bytes; the relevant queue in bytes;
o cq.time() or lq.time() returns the current queuing delay (aka. o cq.time() or lq.time() returns the current queuing delay (aka.
sojourn time or service time) of the relevant queue in units of sojourn time or service time) of the relevant queue in units of
time; time (see Note a);
Queuing delay could be measured directly by storing a per-packet o mark(pkt) and drop(pkt) for ECN-marking and dropping a packet;
time-stamp as each packet is enqueued, and subtracting this from the
system time when the packet is dequeued. If time-stamping is not
easy to introduce with certain hardware, queuing delay could be
predicted indirectly by dividing the size of the queue by the
predicted departure rate, which might be known precisely for some
link technologies (see for example [RFC8034]).
In our experiments so far (building on experiments with PIE) on In experiments so far (building on experiments with PIE) on broadband
broadband access links ranging from 4 Mb/s to 200 Mb/s with base RTTs access links ranging from 4 Mb/s to 200 Mb/s with base RTTs from 5 ms
from 5 ms to 100 ms, DualPI2 achieves good results with the default to 100 ms, DualPI2 achieves good results with the default parameters
parameters in Figure 2. The parameters are categorised by whether in Figure 2. The parameters are categorised by whether they relate
they relate to the Base PI2 AQM, the L4S AQM or the framework to the Base PI2 AQM, the L4S AQM or the framework coupling them
coupling them together. Variables derived from these parameters are together. Constants and variables derived from these parameters are
also included at the end of each category. Each parameter is also included at the end of each category. Each parameter is
explained as it is encountered in the walk-through of the pseudocode explained as it is encountered in the walk-through of the pseudocode
below. below.
1: dualpi2_params_init(...) { % Set input parameter defaults 1: dualpi2_params_init(...) { % Set input parameter defaults
2: % PI2 AQM parameters 2: % DualQ Coupled framework parameters
3: target = 15 ms % PI AQM Classic queue delay target 5: limit = MAX_LINK_RATE * 250 ms % Dual buffer size
4: Tupdate = 16 ms % PI Classic queue sampling interval 3: k = 2 % Coupling factor
5: alpha = 10 Hz^2 % PI integral gain 4: % NOT SHOWN % scheduler-dependent weight or equival't parameter
6: beta = 100 Hz^2 % PI proportional gain 6:
7: p_Cmax = 1/4 % Max Classic drop/mark prob 7: % PI2 AQM parameters
8: % Constants derived from PI2 AQM parameters 8: RTT_max = 100 ms % Worst case RTT expected
9: alpha_U = alpha *Tupdate % PI integral gain per update interval 9: RTT_typ = 15 ms % Typical RTT
10: beta_U = beta * Tupdate % PI prop'nal gain per update interval 11: % PI2 constants derived from above PI2 parameters
11: 10: p_Cmax = min(1/k^2, 1) % Max Classic drop/mark prob
12: % DualQ Coupled framework parameters 12: target = RTT_typ % PI AQM Classic queue delay target
13: k = 2 % Coupling factor 13: Tupdate = min(RTT_typ, RTT_max/3) % PI sampling interval
14: % scheduler weight or equival't parameter (scheduler-dependent) 14: alpha = 0.1 * Tupdate / RTT_max^2 % PI integral gain in Hz
15: limit = MAX_LINK_RATE * 250 ms % Dual buffer size 15: beta = 0.3 / RTT_max % PI proportional gain in Hz
16: 16:
17: % L4S ramp AQM parameters 17: % L4S ramp AQM parameters
18: minTh = 475 us % L4S min marking threshold in time units 18: minTh = 475 us % L4S min marking threshold in time units
19: range = 525 us % Range of L4S ramp in time units 19: range = 525 us % Range of L4S ramp in time units
20: Th_len = 2 * MTU % Min L4S marking threshold in bytes 20: Th_len = 2 * MTU % Min L4S marking threshold in bytes
21: % Constants derived from L4S AQM parameters 21: % L4S constants incl. those derived from other parameters
22: p_Lmax = min(k*sqrt(p_Cmax), 1) % Max L4S marking prob 22: p_Lmax = 1 % Max L4S marking prob
23: floor = Th_len * 8 / MIN_LINK_RATE % MIN_LINK_RATE is in Mb/s 23: floor = Th_len / MIN_LINK_RATE
24: if (minTh < floor) { 24: if (minTh < floor) {
25: % Adjust ramp to exceed serialization time of 2 MTU 25: % Shift ramp so minTh >= serialization time of 2 MTU
26: range = max(range - (floor-minTh), 1) % 1us avoids /0 error 26: minTh = floor
27: minTh = floor 27: }
28: } 28: maxTh = minTh+range % L4S max marking threshold in time units
29: maxTh = minTh+range % L4S min marking threshold in time units 29: }
30: }
Figure 2: Example Header Pseudocode for DualQ Coupled PI2 AQM Figure 2: Example Header Pseudocode for DualQ Coupled PI2 AQM
For brevity the pseudocode shows some parameters in units of The overall goal of the code is to maintain the base probability (p',
microseconds (us), but a real implementation would probably use p-prime as in Section 2.4), which is an internal variable from which
nanoseconds. the marking and dropping probabilities for L4S and Classic traffic
(p_L and p_C) are derived, with p_L in turn being derived from p_CL.
The overall goal of the code is to maintain the base probability (p), The probabilities p_CL and p_C are derived in lines 4 and 5 of the
which is an internal variable from which the marking and dropping dualpi2_update() function (Figure 6) then used in the
probabilities for L4S and Classic traffic (p_L and p_C) are derived. dualpi2_dequeue() function where p_L is also derived from p_CL at
The variable named p in the pseudocode and in this walk-through is line 6 (Figure 4). The code walk-through below builds up to
the same as p' (p-prime) in Section 2.4. The probabilities p_L and explaining that part of the code eventually, but it starts from
p_C are derived in lines 3, 4 and 5 of the dualpi2_update() function packet arrival.
(Figure 6) then used in the dualpi2_dequeue() function (Figure 4).
The code walk-through below builds up to explaining that part of the
code eventually, but it starts from packet arrival.
1: dualpi2_enqueue(lq, cq, pkt) { % Test limit and classify lq or cq 1: dualpi2_enqueue(lq, cq, pkt) { % Test limit and classify lq or cq
2: if ( lq.len() + cq.len() > limit ) 2: if ( lq.len() + cq.len() + MTU > limit)
3: drop(pkt) % drop packet if buffer is full 3: drop(pkt) % drop packet if buffer is full
4: else { % Packet classifier 4: timestamp(pkt) % attach arrival time to packet
5: if ( ecn(pkt) modulo 2 == 1 ) % ECN bits = ECT(1) or CE 5: % Packet classifier
6: lq.enqueue(pkt) 6: if ( ecn(pkt) modulo 2 == 1 ) % ECN bits = ECT(1) or CE
7: else % ECN bits = not-ECT or ECT(0) 7: lq.enqueue(pkt)
8: cq.enqueue(pkt) 8: else % ECN bits = not-ECT or ECT(0)
9: } 9: cq.enqueue(pkt)
10: } 10: }
Figure 3: Example Enqueue Pseudocode for DualQ Coupled PI2 AQM Figure 3: Example Enqueue Pseudocode for DualQ Coupled PI2 AQM
1: dualpi2_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues 1: dualpi2_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: while ( lq.len() + cq.len() > 0 ) 2: while ( lq.len() + cq.len() > 0 )
3: if ( scheduler() == lq ) { 3: if ( scheduler() == lq ) {
4: lq.dequeue(pkt) % Scheduler chooses lq 4: lq.dequeue(pkt) % Scheduler chooses lq
5: p'_L = laqm(lq.time()) % Native L4S AQM 5: p'_L = laqm(lq.time()) % Native L4S AQM
6: p_L = max(p'_L, p_CL) % Combining function 6: p_L = max(p'_L, p_CL) % Combining function
7: if ( p_L > rand() ) % Linear marking 7: if ( recur(p_L) ) % Linear marking
8: mark(pkt) 8: mark(pkt)
9: } else { 9: } else {
10: cq.dequeue(pkt) % Scheduler chooses cq 10: cq.dequeue(pkt) % Scheduler chooses cq
11: if ( p_C > rand() ) { % probability p_C = p^2 11: if ( p_C > rand() ) { % probability p_C = p'^2
12: if ( ecn(pkt) == 0 ) { % if ECN field = not-ECT 12: if ( ecn(pkt) == 0 ) { % if ECN field = not-ECT
13: drop(pkt) % squared drop 13: drop(pkt) % squared drop
14: continue % continue to the top of the while loop 14: continue % continue to the top of the while loop
15: } 15: }
16: mark(pkt) % squared mark 16: mark(pkt) % squared mark
17: } 17: }
18: } 18: }
19: return(pkt) % return the packet and stop 19: return(pkt) % return the packet and stop
20: } 20: }
21: return(NULL) % no packet to dequeue 21: return(NULL) % no packet to dequeue
22: } 22: }
23: recur(likelihood) { % Returns TRUE with a certain likelihood
24: count += likelihood
25: if (count > 1) {
26: count -= 1
27: return TRUE
28: }
29: return FALSE
30: }
Figure 4: Example Dequeue Pseudocode for DualQ Coupled PI2 AQM Figure 4: Example Dequeue Pseudocode for DualQ Coupled PI2 AQM
When packets arrive, first a common queue limit is checked as shown When packets arrive, first a common queue limit is checked as shown
in line 2 of the enqueuing pseudocode in Figure 3. Note that the in line 2 of the enqueuing pseudocode in Figure 3. This assumes a
limit is deliberately tested before enqueue to avoid any bias against shared buffer for the two queues (Note b discusses the merits of
larger packets (so depending whether the implementation stores a separate buffers). In order to avoid any bias against larger
packet while testing whether to drop it from the tail, it might be packets, 1 MTU of space is always allowed and the limit is
necessary for the actual buffer memory to be one MTU larger than deliberately tested before enqueue.
limit).
Line 2 assumes an implementation where lq and cq share common buffer If limit is not exceeded, the packet is timestamped in line 4. This
memory. An alternative implementation could use separate buffers for assumes that queue delay is measured using the sojourn time technique
each queue, in which case the arriving packet would have to be (see Note a for alternatives).
classified first to determine which buffer to check for available
space. The choice is a trade off; a shared buffer can use less
memory whereas separate buffers isolate the L4S queue from tail-drop
due to large bursts of Classic traffic (e.g. a Classic TCP during
slow-start over a long RTT).
Returning to the shared buffer case, if limit is not exceeded, the At lines 5-9, the packet is classified and enqueued to the Classic or
packet will be classified and enqueued to the Classic or L4S queue L4S queue dependent on the least significant bit of the ECN field in
dependent on the least significant bit of the ECN field in the IP the IP header (line 6). Packets with a codepoint having an LSB of 0
header (line 5). Packets with a codepoint having an LSB of 0 (Not- (Not-ECT and ECT(0)) will be enqueued in the Classic queue.
ECT and ECT(0)) will be enqueued in the Classic queue. Otherwise, Otherwise, ECT(1) and CE packets will be enqueued in the L4S queue.
ECT(1) and CE packets will be enqueued in the L4S queue. Optional Optional additional packet classification flexibility is omitted for
additional packet classification flexibility is omitted for brevity brevity (see [I-D.ietf-tsvwg-ecn-l4s-id]).
(see [I-D.ietf-tsvwg-ecn-l4s-id]).
The dequeue pseudocode (Figure 4) is repeatedly called whenever the The dequeue pseudocode (Figure 4) is repeatedly called whenever the
lower layer is ready to forward a packet. It schedules one packet lower layer is ready to forward a packet. It schedules one packet
for dequeuing (or zero if the queue is empty) then returns control to for dequeuing (or zero if the queue is empty) then returns control to
the caller, so that it does not block while that packet is being the caller, so that it does not block while that packet is being
forwarded. While making this dequeue decision, it also makes the forwarded. While making this dequeue decision, it also makes the
necessary AQM decisions on dropping or marking. The alternative of necessary AQM decisions on dropping or marking. The alternative of
applying the AQMs at enqueue would shift some processing from the applying the AQMs at enqueue would shift some processing from the
critical time when each packet is dequeued. However, it would also critical time when each packet is dequeued. However, it would also
add a whole queue of delay to the control signals, making the control add a whole queue of delay to the control signals, making the control
loop very sloppy. loop sloppier (for a typical RTT it would double the Classic queue's
feedback delay).
All the dequeue code is contained within a large while loop so that All the dequeue code is contained within a large while loop so that
if it decides to drop a packet, it will continue until it selects a if it decides to drop a packet, it will continue until it selects a
packet to schedule. Line 3 of the dequeue pseudocode is where the packet to schedule. Line 3 of the dequeue pseudocode is where the
scheduler chooses between the L4S queue (lq) and the Classic queue scheduler chooses between the L4S queue (lq) and the Classic queue
(cq). Detailed implementation of the scheduler is not shown (see (cq). Detailed implementation of the scheduler is not shown (see
discussion later). discussion later).
o If an L4S packet is scheduled, lines 7 and 8 ECN-mark the packet o If an L4S packet is scheduled, in lines 7 and 8 the packet is ECN-
if a random marking decision is drawn according to p_L. Line 6 marked with likelihood p_L. The recur() function at the end of
calculates p_L as the maximum of the coupled L4S probability p_CL Figure 4 is used, which is preferred over random marking because
and the probability from the native L4S AQM p'_L. This implements it avoids delay due to randomization when interpreting congestion
the max() function shown in Figure 1 to couple the outputs of the signals, but it still desynchronizes the saw-teeth of the flows.
two AQMs together. Of the two probabilities input to p_L in line Line 6 calculates p_L as the maximum of the coupled L4S
6: probability p_CL and the probability from the native L4S AQM p'_L.
This implements the max() function shown in Figure 1 to couple the
outputs of the two AQMs together. Of the two probabilities input
to p_L in line 6:
* p'_L is calculated per packet in line 5 by the laqm() function * p'_L is calculated per packet in line 5 by the laqm() function
(see Figure 5), (see Figure 5),
* whereas p_CL is maintained by the dualpi2_update() function * whereas p_CL is maintained by the dualpi2_update() function
which runs every Tupdate (default 16ms) (see Figure 2). which runs every Tupdate (Tupdate is set in line 13 of
Figure 2. It defaults to 16 ms in the reference Linux
implementation because it has to be rounded to a multiple of 4
ms).
o If a Classic packet is scheduled, lines 10 to 17 drop or mark the o If a Classic packet is scheduled, lines 10 to 17 drop or mark the
packet based on the squared probability p_C. packet with probability p_C.
The Native L4S AQM algorithm (Figure 5) is a ramp function, similar The Native L4S AQM algorithm (Figure 5) is a ramp function, similar
to the RED algorithm, but simpler due to the following differences: to the RED algorithm, but simplified as follows:
o The min and max of the ramp are defined in units of queuing delay, o The extent of the ramp is defined in units of queuing delay, not
not bytes, so that configuration remains invariant as the queue bytes, so that configuration remains invariant as the queue
departure rate varies. departure rate varies.
o It uses instantaneous queueing delay to remove smoothing delay o It uses instantaneous queueing delay, which avoids the complexity
(L4S senders smooth incoming ECN feedback when necessary). of smoothing, but also avoids embedding a worst-case RTT of
smoothing delay in the network (see Section 2.1).
o The ramp rises linearly directly from 0 to 1, not to a an o The ramp rises linearly directly from 0 to 1, not to a an
intermediate value of p'_L as RED would, because there is no need intermediate value of p'_L as RED would, because there is no need
to keep ECN marking probability low. to keep ECN marking probability low.
o Marking does not have to be randomized. Determinism is being o Marking does not have to be randomized. Determinism is used
experimented with instead of randomness; to reduce the delay instead of randomness; to reduce the delay necessary to smooth out
necessary to smooth out the noise of randomness from the signal. the noise of randomness from the signal.
In this case, for each packet, the algorithm would accumulate p_L
in a counter and mark the packet that took the counter over 1,
then subtract 1 from the counter and continue.
This ramp function requires two configuration parameters, the minimum The ramp function requires two configuration parameters, the minimum
threshold (minTh) and the width of the ramp (range), both in units of threshold (minTh) and the width of the ramp (range), both in units of
queuing time), as shown in the parameter initialization code in queuing time), as shown in lines 18 & 19 of the initialization
Figure 2. A minimum marking threshold parameter (Th_len) in function in Figure 2. The ramp function can be configured as a step
transmission units (default 2 MTU) is also necessary to ensure that (see Note c).
the ramp does not trigger excessive marking on slow links. The code
in lines 23-28 of Figure 2 converts 2 MTU into time units and adjusts
the ramp thresholds to be no shallower than this floor.
An operator can effectively turn the ramp into a step function, as Although the DCTCP paper [Alizadeh-stability] recommends an ECN
used by DCTCP, by setting the range to its minimum value (e.g. 1 ns). marking threshold of 0.17*RTT_typ, it also shows that the threshold
Then the condition for the ramp calculation will hardly ever arise. can be much shallower with hardly any worse under-utilization of the
There is some concern that using the step function of DCTCP for the link (because the amplitude of DCTCP's sawteeth is so small). Based
Native L4S AQM requires end-systems to smooth the signal for an on extensive experiments, for the public Internet a default minimum
unnecessarily large number of round trips to ensure sufficient ECN marking threshold of about RTT_typ/30 is recommended.
fidelity. A ramp seems to be no worse than a step in initial
experiments with existing DCTCP. Therefore, it is recommended that a A minimum marking threshold parameter (Th_len) in transmission units
ramp is configured in place of a step, which will allow congestion (default 2 MTU) is also necessary to ensure that the ramp does not
control algorithms to investigate faster smoothing algorithms. trigger excessive marking on slow links. The code in lines 24-27 of
the initialization function (Figure 2) converts 2 MTU into time units
and shifts the ramp so that the min threshold is no shallower than
this floor.
1: laqm(qdelay) { % Returns native L4S AQM probability 1: laqm(qdelay) { % Returns native L4S AQM probability
2: if (qdelay >= maxTh) 2: if (qdelay >= maxTh)
3: return 1 3: return 1
4: else if (qdelay > minTh) 4: else if (qdelay > minTh)
5: return (qdelay - minTh)/range % Divide would use a bit-shift 5: return (qdelay - minTh)/range % Divide could use a bit-shift
6: else 6: else
7: return 0 7: return 0
8: } 8: }
Figure 5: Example Pseudocode for the Native L4S AQM Figure 5: Example Pseudocode for the Native L4S AQM
1: dualpi2_update(lq, cq, target) { % Update p every Tupdate 1: dualpi2_update(lq, cq) { % Update p' every Tupdate
2: curq = cq.time() % use queuing time of first-in Classic packet 2: curq = cq.time() % use queuing time of first-in Classic packet
3: p = p + alpha_U * (curq - target) + beta_U * (curq - prevq) 3: p' = p' + alpha * (curq - target) + beta * (curq - prevq)
4: p_CL = p * k % Coupled L4S prob = base prob * coupling factor 4: p_CL = k * p' % Coupled L4S prob = base prob * coupling factor
5: p_C = p^2 % Classic prob = (base prob)^2 5: p_C = p'^2 % Classic prob = (base prob)^2
6: prevq = curq 6: prevq = curq
7: } 7: }
Figure 6: Example PI-Update Pseudocode for DualQ Coupled PI2 AQM Figure 6: Example PI-Update Pseudocode for DualQ Coupled PI2 AQM
p_CL depends on the base probability (p), which is kept up to date by The coupled marking probability, p_CL depends on the base probability
the core PI algorithm in Figure 6 executed every Tupdate. (p'), which is kept up to date by the core PI algorithm in Figure 6
executed every Tupdate.
Note that p solely depends on the queuing time in the Classic queue. Note that p' solely depends on the queuing time in the Classic queue.
In line 2, the current queuing delay (curq) is evaluated from how In line 2, the current queuing delay (curq) is evaluated from how
long the head packet was in the Classic queue (cq). The function long the head packet was in the Classic queue (cq). The function
cq.time() (not shown) subtracts the time stamped at enqueue from the cq.time() (not shown) subtracts the time stamped at enqueue from the
current time and implicitly takes the current queuing delay as 0 if current time (see Note a) and implicitly takes the current queuing
the queue is empty. delay as 0 if the queue is empty.
The algorithm centres on line 3, which is a classical Proportional- The algorithm centres on line 3, which is a classical Proportional-
Integral (PI) controller that alters p dependent on: a) the error Integral (PI) controller that alters p' dependent on: a) the error
between the current queuing delay (curq) and the target queuing delay between the current queuing delay (curq) and the target queuing delay
('target' - see [RFC8033]); and b) the change in queuing delay since ('target' - see [RFC8033]); and b) the change in queuing delay since
the last sample. The name 'PI' represents the fact that the second the last sample. The name 'PI' represents the fact that the second
factor (how fast the queue is growing) is _P_roportional to load factor (how fast the queue is growing) is _P_roportional to load
while the first is the _I_ntegral of the load (so it removes any while the first is the _I_ntegral of the load (so it removes any
standing queue in excess of the target). standing queue in excess of the target).
The two 'gain factors' in line 3, alpha_U and beta_U, respectively The two 'gain factors' in line 3, alpha and beta, respectively weight
weight how strongly each of these elements ((a) and (b)) alters p. how strongly each of these elements ((a) and (b)) alters p'. They
They are in units of 'per second of delay' or Hz, because they are in units of 'per second of delay' or Hz, because they transform
transform differences in queueing delay into changes in probability. differences in queueing delay into changes in probability (assuming
probability has a value from 0 to 1).
alpha_U and beta_U are derived from the input parameters alpha and alpha and beta determine how much p' ought to change after each
beta (see lines 5 and 6 of Figure 2). These recommended values of update interval (Tupdate). For smaller Tupdate, p' should change by
alpha and beta come from the stability analysis in [PI2] so that the the same amount per second, but in finer more frequent steps. So
AQM can change p as fast as possible in response to changes in load alpha depends on Tupdate (see line 14 of the initialization function
without over-compensating and therefore causing oscillations in the in Figure 2). It is best to update p' as frequently as possible, but
queue. Tupdate will probably be constrained by hardware performance. As
shown in line 13, the update interval should be at least as frequent
as once per the RTT of a typical flow (RTT_typ) as long as it does
not exceed roughly RTT_max/3. For link rates from 4 - 200 Mb/s, a
target RTT of 15ms and a maximum RTT of 100ms, it has been verified
through extensive testing that Tupdate=16ms (as recommended in
[RFC8033]) is sufficient.
alpha and beta determine how much p ought to change if it was updated The choice of alpha and beta also determines the AQM's stable
every second. It is best to update p as frequently as possible, but operating range. The AQM ought to change p' as fast as possible in
the update interval (Tupdate) will probably be constrained by response to changes in load without over-compensating and therefore
hardware performance. For link rates from 4 - 200 Mb/s, we found causing oscillations in the queue. Therefore, the values of alpha
Tupdate=16ms (as recommended in [RFC8033]) is sufficient. However and beta also depend on the RTT of the expected worst-case flow
small the chosen value of Tupdate, p should change by the same amount (RTT_max).
per second, but in finer more frequent steps. So the gain factors
used for updating p in Figure 6 need to be scaled by (Tupdate/1s),
which is done in lines 9 and 10 of Figure 2). The suffix '_U'
represents 'per update time' (Tupdate).
In corner cases, p can overflow the range [0,1] so the resulting Recommended derivations of the gain constants alpha and beta can be
value of p has to be bounded (omitted from the pseudocode). Then, as approximated for Reno over a PI2 AQM as: alpha = 0.1 * Tupdate /
already explained, the coupled and Classic probabilities are derived RTT_max^2; beta = 0.3 / RTT_max, as shown in lines 14 & 15 of
from the new p in lines 4 and 5 as p_CL = k*p and p_C = p^2. Figure 2. These are derived from the stability analysis in [PI2].
For the default values of Tupdate=16 ms and RTT_max = 100 ms, they
result in alpha = 0.16; beta = 3.2 (discrepancies are due to
rounding). These defaults have been verified with a wide range of
link rates, target delays and a range of traffic models with mixed
and similar RTTs, short and long flows, etc.
In corner cases, p' can overflow the range [0,1] so the resulting
value of p' has to be bounded (omitted from the pseudocode). Then,
as already explained, the coupled and Classic probabilities are
derived from the new p' in lines 4 and 5 of Figure 6 as p_CL = k*p'
and p_C = p'^2.
Because the coupled L4S marking probability (p_CL) is factored up by Because the coupled L4S marking probability (p_CL) is factored up by
k, the dynamic gain parameters alpha and beta are also inherently k, the dynamic gain parameters alpha and beta are also inherently
factored up by k for the L4S queue, which is necessary to ensure that factored up by k for the L4S queue. So, the effective gain factor
Classic TCP and DCTCP controls have the same stability. So, if alpha for the L4S queue is k*alpha (with defaults alpha = 0.16 Hz and k=2,
is 10 Hz^2, the effective gain factor for the L4S queue is k*alpha, effective L4S alpha = 0.32 Hz).
which is 20 Hz^2 with the default coupling factor of k=2.
Unlike in PIE [RFC8033], alpha_U and beta_U do not need to be tuned Unlike in PIE [RFC8033], alpha and beta do not need to be tuned every
every Tupdate dependent on p. Instead, in PI2, alpha_U and beta_U Tupdate dependent on p'. Instead, in PI2, alpha and beta are
are independent of p because the squaring applied to Classic traffic independent of p' because the squaring applied to Classic traffic
tunes them inherently. This is explained in [PI2], which also tunes them inherently. This is explained in [PI2], which also
explains why this more principled approach removes the need for most explains why this more principled approach removes the need for most
of the heuristics that had to be added to PIE. of the heuristics that had to be added to PIE.
{ToDo: Scaling beta with Tupdate and scaling both alpha & beta with Notes:
RTT}
a. The drain rate of the queue can vary if it is scheduled relative
to other queues, or to cater for fluctuations in a wireless
medium. To auto-adjust to changes in drain rate, the queue must
be measured in time, not bytes or packets [AQMmetrics] [CoDel].
Queuing delay could be measured directly by storing a per-packet
time-stamp as each packet is enqueued, and subtracting this from
the system time when the packet is dequeued. If time-stamping is
not easy to introduce with certain hardware, queuing delay could
be predicted indirectly by dividing the size of the queue by the
predicted departure rate, which might be known precisely for some
link technologies (see for example [RFC8034]).
b. Line 2 of the dualpi2_enqueue() function (Figure 3) assumes an
implementation where lq and cq share common buffer memory. An
alternative implementation could use separate buffers for each
queue, in which case the arriving packet would have to be
classified first to determine which buffer to check for available
space. The choice is a trade off; a shared buffer can use less
memory whereas separate buffers isolate the L4S queue from tail-
drop due to large bursts of Classic traffic (e.g. a Classic TCP
during slow-start over a long RTT).
c. There has been some concern that using the step function of DCTCP
for the Native L4S AQM requires end-systems to smooth the signal
for an unnecessarily large number of round trips to ensure
sufficient fidelity. A ramp is no worse than a step in initial
experiments with existing DCTCP. Therefore, it is recommended
that a ramp is configured in place of a step, which will allow
congestion control algorithms to investigate faster smoothing
algorithms.
A ramp is more general that a step, because an operator can
effectively turn the ramp into a step function, as used by DCTCP,
by setting the range to zero. There will not be a divide by zero
problem at line 4 of Figure 5 because, if minTh is equal to
maxTh, the condition for this ramp calculation cannot arise.
A.2. Pass #2: Overload Details A.2. Pass #2: Overload Details
Figure 7 repeats the dequeue function of Figure 4, but with overload Figure 7 repeats the dequeue function of Figure 4, but with overload
details added. Similarly Figure 8 repeats the core PI algorithm of details added. Similarly Figure 8 repeats the core PI algorithm of
Figure 6 with overload details added. The initialization, enqueue Figure 6 with overload details added. The initialization, enqueue,
and L4S AQM functions are unchanged. L4S AQM and recur functions are unchanged.
In line 7 of the initialization function (Figure 2), the default In line 10 of the initialization function (Figure 2), the maximum
maximum Classic drop probability p_Cmax = 1/4 or 25%. This is the Classic drop probability p_Cmax = min(1/k^2, 1) or 1/4 for the
point at which it is deemed that the Classic queue has become default coupling factor k=2. p_Cmax is the point at which it is
persistently overloaded, so it switches to using solely drop, even deemed that the Classic queue has become persistently overloaded, so
for ECN-capable packets. This protects the queue against any it switches to using drop, even for ECN-capable packets. ECT packets
unresponsive traffic that falsely claims that it is responsive to ECN that are not dropped can still be ECN-marked.
marking, as required by [RFC3168] and [RFC7567].
Line 22 of the initialization function translates this into a maximum In practice, 25% has been found to be a good threshold to preserve
L4S marking probability (p_Lmax) by rearranging Equation (1). With a fairness between ECN capable and non ECN capable traffic. This
coupling factor of k=2 (the default) or greater, this translates to a protects the queues against both temporary overload from responsive
maximum L4S marking probability of 1 (or 100%). This is intended to flows and more persistent overload from any unresponsive traffic that
ensure that the L4S queue starts to introduce dropping once marking falsely claims to be responsive to ECN.
saturates and can rise no further. The 'TCP Prague' requirements
When the Classic ECN marking probability reaches the p_Cmax threshold
(1/k^2), the marking probability coupled to the L4S queue, p_CL will
always be 100% for any k (by equation (1) in Section 2). So, for
readability, the constant p_Lmax is defined as 1 in line 22 of the
initialization function Figure 2. This is intended to ensure that
the L4S queue starts to introduce dropping once ECN-marking saturates
at 100% and can rise no further. The 'Prague L4S' requirements
[I-D.ietf-tsvwg-ecn-l4s-id] state that, when an L4S congestion [I-D.ietf-tsvwg-ecn-l4s-id] state that, when an L4S congestion
control detects a drop, it falls back to a response that coexists control detects a drop, it falls back to a response that coexists
with 'Classic' TCP. So it is correct that the L4S queue drops with 'Classic' TCP. So it is correct that, when the L4S queue drops
packets proportional to p^2, as if they are Classic packets. packets, it drops them proportional to p'^2, as if they are Classic
packets.
Both these switch-overs are triggered by the tests for overload Both these switch-overs are triggered by the tests for overload
introduced in lines 4b and 12b of the dequeue function (Figure 7). introduced in lines 4b and 12b of the dequeue function (Figure 7).
Lines 8c to 8g drop L4S packets with probability p^2. Lines 8h to 8i Lines 8c to 8g drop L4S packets with probability p'^2. Lines 8h to
mark the remaining packets with probability p_CL. If p_Lmax = 1, 8i mark the remaining packets with probability p_CL. Given p_Lmax =
which is the suggested default configuration, all remaining packets 1, all remaining packets will be marked because, to have reached the
will be marked because, to have reached the else block at line 8b, else block at line 8b, p_CL >= 1.
p_CL >= 1.
Lines 2c to 2d in the core PI algorithm (Figure 8) deal with overload Lines 2c to 2d in the core PI algorithm (Figure 8) deal with overload
of the L4S queue when there is no Classic traffic. This is of the L4S queue when there is no Classic traffic. This is
necessary, because the core PI algorithm maintains the appropriate necessary, because the core PI algorithm maintains the appropriate
drop probability to regulate overload, but it depends on the length drop probability to regulate overload, but it depends on the length
of the Classic queue. If there is no Classic queue the naive of the Classic queue. If there is no Classic queue the naive PI
algorithm in Figure 6 drops nothing, even if the L4S queue is update function in Figure 6 would drop nothing, even if the L4S queue
overloaded - so tail drop would have to take over (lines 3 and 4 of were overloaded - so tail drop would have to take over (lines 2 and 3
Figure 3). of Figure 3).
If the test at line 2a finds that the Classic queue is empty, line 2d Instead, the test at line 2a of the full PI update function in
measures the current queue delay using the L4S queue instead. While Figure 8 keeps delay on target using drop. If the test at line 2a of
the L4S queue is not overloaded, its delay will always be tiny finds that the Classic queue is empty, line 2d measures the current
compared to the target Classic queue delay. So p_L will be driven to queue delay using the L4S queue instead. While the L4S queue is not
zero, and the L4S queue will naturally be governed solely by overloaded, its delay will always be tiny compared to the target
threshold marking (lines 5 and 6 of the dequeue algorithm in Classic queue delay. So p_CL will be driven to zero, and the L4S
Figure 7). But, if unresponsive L4S source(s) cause overload, the queue will naturally be governed solely by p'_L from the native L4S
DualQ transitions smoothly to L4S marking based on the PI algorithm. AQM (lines 5 and 6 of the dequeue algorithm in Figure 7). But, if
And as overload increases, it naturally transitions from marking to unresponsive L4S source(s) cause overload, the DualQ transitions
dropping by the switch-over mechanism already described. smoothly to L4S marking based on the PI algorithm. If overload
increases further, it naturally transitions from marking to dropping
by the switch-over mechanism already described.
1: dualpi2_dequeue(lq, cq) { % Couples L4S & Classic queues, lq & cq 1: dualpi2_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: while ( lq.len() + cq.len() > 0 ) 2: while ( lq.len() + cq.len() > 0 ) {
3: if ( scheduler() == lq ) { 3: if ( scheduler() == lq ) {
4a: lq.dequeue(pkt) 4a: lq.dequeue(pkt) % L4S scheduled
4b: if ( p_CL < p_Lmax ) { % Check for overload saturation 4b: if ( p_CL < p_Lmax ) { % Check for overload saturation
5: p'_L = laqm(lq.time()) % Native L4S AQM 5: p'_L = laqm(lq.time()) % Native L4S AQM
6: p_L = max(p'_L, p_CL) % Combining function 6: p_L = max(p'_L, p_CL) % Combining function
7: if ( p_L > rand() ) % Linear marking 7: if ( recur(p_L) ) % Linear marking
8a: mark(pkt) 8a: mark(pkt)
8b: } else { % overload saturation 8b: } else { % overload saturation
8c: if ( p_C > rand() ) { % probability p_C = p^2 8c: if ( p_C > rand() ) { % probability p_C = p'^2
8e: drop(pkt) % revert to Classic drop due to overload 8e: drop(pkt) % revert to Classic drop due to overload
8f: continue % continue to the top of the while loop 8f: continue % continue to the top of the while loop
8g: } 8g: }
8h: if ( p_CL > rand() ) % probability p_CL = k * p 8h: if ( p_CL > rand() ) % probability p_CL = k * p'
8i: mark(pkt) % linear marking of remaining packets 8i: mark(pkt) % linear marking of remaining packets
8j: } 8j: }
9: } else { 9: } else {
10: cq.dequeue(pkt) 10: cq.dequeue(pkt) % Classic scheduled
11: if ( p_C > rand() ) { % probability p_C = p^2 11: if ( p_C > rand() ) { % probability p_C = p'^2
12a: if ( (ecn(pkt) == 0) % ECN field = not-ECT 12a: if ( (ecn(pkt) == 0) % ECN field = not-ECT
12b: OR (p_C >= p_Cmax) ) { % Overload disables ECN 12b: OR (p_C >= p_Cmax) ) { % Overload disables ECN
13: drop(pkt) % squared drop, redo loop 13: drop(pkt) % squared drop, redo loop
14: continue % continue to the top of the while loop 14: continue % continue to the top of the while loop
15: } 15: }
16: mark(pkt) % squared mark 16: mark(pkt) % squared mark
17: } 17: }
18: } 18: }
19: return(pkt) % return the packet and stop 19: return(pkt) % return the packet and stop
20: } 20: }
21: return(NULL) % no packet to dequeue 21: return(NULL) % no packet to dequeue
22: } 22: }
Figure 7: Example Dequeue Pseudocode for DualQ Coupled PI2 AQM Figure 7: Example Dequeue Pseudocode for DualQ Coupled PI2 AQM
(Including Integer Arithmetic and Overload Code) (Including Overload Code)
1: dualpi2_update(lq, cq, target) { % Update p every Tupdate 1: dualpi2_update(lq, cq) { % Update p' every Tupdate
2a: if ( cq.len() > 0 ) 2a: if ( cq.len() > 0 )
2b: curq = cq.time() %use queuing time of first-in Classic packet 2b: curq = cq.time() %use queuing time of first-in Classic packet
2c: else % Classic queue empty 2c: else % Classic queue empty
2d: curq = lq.time() % use queuing time of first-in L4S packet 2d: curq = lq.time() % use queuing time of first-in L4S packet
3: p = p + alpha_U * (curq - target) + beta_U * (curq - prevq) 3: p' = p' + alpha * (curq - target) + beta * (curq - prevq)
4: p_CL = p * k % Coupled L4S prob = base prob * coupling factor 4: p_CL = p' * k % Coupled L4S prob = base prob * coupling factor
5: p_C = p^2 % Classic prob = (base prob)^2 5: p_C = p'^2 % Classic prob = (base prob)^2
6: prevq = curq 6: prevq = curq
7: } 7: }
Figure 8: Example PI-Update Pseudocode for DualQ Coupled PI2 AQM Figure 8: Example PI-Update Pseudocode for DualQ Coupled PI2 AQM
(Including Overload Code) (Including Overload Code)
The choice of scheduler technology is critical to overload protection The choice of scheduler technology is critical to overload protection
(see Section 4.1). (see Section 4.1).
o A well-understood weighted scheduler such as weighted round robin o A well-understood weighted scheduler such as weighted round robin
(WRR) is recommended. The scheduler weight for Classic should be (WRR) is recommended. As long as the scheduler weight for Classic
low, e.g. 1/16. is small (e.g. 1/16), its exact value is unimportant because it
does not normally determine capacity shares. The weight is only
important to prevent unresponsive L4S traffic starving Classic
traffic. This is because capacity sharing between the queues is
normally determined by the coupled congestion signal, which
overrides the scheduler, by making L4S sources leave roughly equal
per-flow capacity available for Classic flows.
o Alternatively, a time-shifted FIFO could be used. This is a very o Alternatively, a time-shifted FIFO (TS-FIFO) could be used. It
simple scheduler, but it does not fully isolate latency in the L4S works by selecting the head packet that has waited the longest,
queue from uncontrolled bursts in the Classic queue. It works by biased against the Classic traffic by a time-shift of tshift. To
selecting the head packet that has waited the longest, biased implement time-shifted FIFO, the scheduler() function in line 3 of
against the Classic traffic by a time-shift of tshift. To the dequeue code would simply be implemented as the scheduler()
implement time-shifted FIFO, the "if (scheduler() == lq )" test in function at the bottom of Figure 10 in Appendix B. For the public
line 3 of the dequeue code would simply be replaced by "if ( Internet a good value for tshift is 50ms. For private networks
lq.time() + tshift >= cq.time() )". For the public Internet a with smaller diameter, about 4*target would be reasonable. TS-
good value for tshift is 50ms. For private networks with smaller FIFO is a very simple scheduler, but complexity might need to be
diameter, about 4*target would be reasonable. added to address some deficiencies (which is why it is not
recommended over WRR):
* TS-FIFO does not fully isolate latency in the L4S queue from
uncontrolled bursts in the Classic queue;
* TS-FIFO is only appropriate if time-stamping of packets is
feasible;
* Even if time-stamping is supported, the sojourn time of the
head packet is always stale. For instance, if a burst arrives
at an empty queue, the sojourn time will only measure the delay
of the burst once the burst is over, even though the queue knew
about it from the start. At the cost of more operations and
more storage, a 'scaled sojourn time' metric of queue delay can
be used, which is the sojourn time of a packet scaled by the
ratio of the queue sizes when the packet departed and arrived
[SigQ-Dyn].
o A strict priority scheduler would be inappropriate, because it o A strict priority scheduler would be inappropriate, because it
would starve Classic if L4S was overloaded. would starve Classic if L4S was overloaded.
Appendix B. Example DualQ Coupled Curvy RED Algorithm Appendix B. Example DualQ Coupled Curvy RED Algorithm
As another example of a DualQ Coupled AQM algorithm, the pseudocode As another example of a DualQ Coupled AQM algorithm, the pseudocode
below gives the Curvy RED based algorithm we used and tested. below gives the Curvy RED based algorithm. Although the AQM was
Although we designed the AQM to be efficient in integer arithmetic, designed to be efficient in integer arithmetic, to aid understanding
to aid understanding it is first given using real-number arithmetic. it is first given using floating point arithmetic (Figure 10). Then,
Then, one possible optimization for integer arithmetic is given, also one possible optimization for integer arithmetic is given, also in
in pseudocode. To aid comparison, the line numbers are kept in step pseudocode (Figure 11). To aid comparison, the line numbers are kept
between the two by using letter suffixes where the longer code needs in step between the two by using letter suffixes where the longer
extra lines. code needs extra lines.
1: dualq_dequeue(lq, cq) { % Couples L4S & Classic queues, lq & cq B.1. Curvy RED in Pseudocode
2: if ( lq.dequeue(pkt) ) {
3a: p_L = cq.sec() / 2^S_L
3b: if ( lq.byt() > T )
3c: mark(pkt)
3d: elif ( p_L > maxrand(U) )
4: mark(pkt)
5: return(pkt) % return the packet and stop here
6: }
7: while ( cq.dequeue(pkt) ) {
8a: alpha = 2^(-f_C)
8b: Q_C = alpha * pkt.sec() + (1-alpha)* Q_C % Classic Q EWMA
9a: sqrt_p_C = Q_C / 2^S_C
9b: if ( sqrt_p_C > maxrand(2*U) )
10: drop(pkt) % Squared drop, redo loop
11: else
12: return(pkt) % return the packet and stop here
13: }
14: return(NULL) % no packet to dequeue
15: }
16: maxrand(u) { % return the max of u random numbers The pseudocode manipulates three main structures of variables: the
17: maxr=0 packet (pkt), the L4S queue (lq) and the Classic queue (cq) and
18: while (u-- > 0) consists of the following five functions:
19: maxr = max(maxr, rand()) % 0 <= rand() < 1
20: return(maxr) o the initialization function cred_params_init(...) (Figure 2) that
sets parameter defaults (the API for setting non-default values is
omitted for brevity);
o the dequeue function cred_dequeue(lq, cq, pkt) (Figure 4);
o the scheduling function scheduler(), which selects between the
head packets of the two queues.
It also uses the following functions that are either shown elsewhere,
or not shown in full here:
o the enqueue function, which is identical to that used for DualPI2,
dualpi2_enqueue(lq, cq, pkt) in Figure 3;
o mark(pkt) and drop(pkt) for ECN-marking and dropping a packet;
o cq.len() or lq.len() returns the current length (aka. backlog) of
the relevant queue in bytes;
o cq.time() or lq.time() returns the current queuing delay (aka.
sojourn time or service time) of the relevant queue in units of
time (see Note a in Appendix A.1).
Because Curvy RED was evaluated before DualPI2, certain improvements
introduced for DualPI2 were not evaluated for Curvy RED. In the
pseudocode below, the straightforward improvements have been added on
the assumption they will provide similar benefits, but that has not
been proven experimentally. They are: i) a conditional priority
scheduler instead of strict priority ii) a time-based threshold for
the native L4S AQM; iii) ECN support for the Classic AQM. A recent
evaluation has proved that a minimum ECN-marking threshold (minTh)
greatly improves performance, so this is also included in the
pseudocode.
Overload protection has not been added to the Curvy RED pseudocode
below so as not to detract from the main features. It would be added
in exactly the same way as in Appendix A.2 for the DualPI2
pseudocode. The native L4S AQM uses a step threshold, but a ramp
like that described for DualPI2 could be used instead. The scheduler
uses the simple TS-FIFO algorithm, but it could be replaced with WRR.
The Curvy RED algorithm has not been maintained or evaluated to the
same degree as the DualPI2 algorithm. In initial experiments on
broadband access links ranging from 4 Mb/s to 200 Mb/s with base RTTs
from 5 ms to 100 ms, Curvy RED achieved good results with the default
parameters in Figure 9.
The parameters are categorised by whether they relate to the Classic
AQM, the L4S AQM or the framework coupling them together. Constants
and variables derived from these parameters are also included at the
end of each category. These are the raw input parameters for the
algorithm. A configuration front-end could accept more meaningful
parameters (e.g. RTT_max and RTT_typ) and convert them into these
raw parameters, as has been done for DualPI2 in Appendix A. Where
necessary, parameters are explained further in the walk-through of
the pseudocode below.
1: cred_params_init(...) { % Set input parameter defaults
2: % DualQ Coupled framework parameters
3: limit = MAX_LINK_RATE * 250 ms % Dual buffer size
4: k' = 1 % Coupling factor as a power of 2
5: tshift = 50 ms % Time shift of TS-FIFO scheduler
6: % Constants derived from Classic AQM parameters
7: k = 2^k' % Coupling factor from Equation (1)
6:
7: % Classic AQM parameters
8: g_C = 5 % EWMA smoothing parameter as a power of 1/2
9: S_C = -1 % Classic ramp scaling factor as a power of 2
10: minTh = 500 ms % No Classic drop/mark below this queue delay
11: % Constants derived from Classic AQM parameters
12: gamma = 2^(-g_C) % EWMA smoothing parameter
13: range_C = 2^S_C % Range of Classic ramp
14:
15: % L4S AQM parameters
16: T = 1 ms % Queue delay threshold for native L4S AQM
17: % Constants derived from above parameters
18: S_L = S_C - k' % L4S ramp scaling factor as a power of 2
19: range_L = 2^S_L % Range of L4S ramp
20: }
Figure 9: Example Header Pseudocode for DualQ Coupled Curvy RED AQM
1: cred_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: while ( lq.len() + cq.len() > 0 ) {
3: if ( scheduler() == lq ) {
4: lq.dequeue(pkt) % L4S scheduled
5a: p_CL = (cq.time() - minTh) / range_L
5b: if ( ( lq.time() > T )
5c: OR ( p_CL > maxrand(U) ) )
6: mark(pkt)
7: } else {
8: cq.dequeue(pkt) % Classic scheduled
9a: Q_C = gamma * qc.time() + (1-gamma) * Q_C % Classic Q EWMA
10a: sqrt_p_C = (Q_C - minTh) / range_C
10b: if ( sqrt_p_C > maxrand(2*U) ) {
11: if ( (ecn(pkt) == 0) { % ECN field = not-ECT
12: drop(pkt) % Squared drop, redo loop
13: continue % continue to the top of the while loop
14: }
15: mark(pkt)
16: }
17: }
18: return(pkt) % return the packet and stop here
19: }
20: return(NULL) % no packet to dequeue
21: } 21: }
Figure 9: Example Dequeue Pseudocode for DualQ Coupled Curvy RED AQM 22: maxrand(u) { % return the max of u random numbers
23: maxr=0
24: while (u-- > 0)
25: maxr = max(maxr, rand()) % 0 <= rand() < 1
26: return(maxr)
27: }
Packet classification code is not shown, as it is no different from 28: scheduler() {
Figure 3. Potential classification schemes are discussed in 29: if ( lq.time() + tshift >= cq.time() )
Section 2.3. The Curvy RED algorithm has not been maintained to the 30: return lq;
same degree as the DualPI2 algorithm. Some ideas used in DualPI2 31: else
would need to be translated into Curvy RED, such as i) the 32: return cq;
conditional priority scheduler instead of strict priority ii) the 33: }
time-based L4S threshold; iii) turning off ECN as overload
protection; iv) Classic ECN support. These are not shown in the
Curvy RED pseudocode, but would need to be implemented for
production. {ToDo}
At the outer level, the structure of dualq_dequeue() implements Figure 10: Example Dequeue Pseudocode for DualQ Coupled Curvy RED AQM
strict priority scheduling. The code is written assuming the AQM is
applied on dequeue (Note 1) . Every time dualq_dequeue() is called,
the if-block in lines 2-6 determines whether there is an L4S packet
to dequeue by calling lq.dequeue(pkt), and otherwise the while-block
in lines 7-13 determines whether there is a Classic packet to
dequeue, by calling cq.dequeue(pkt). (Note 2)
In the lower priority Classic queue, a while loop is used so that, if
the AQM determines that a classic packet should be dropped, it
continues to test for classic packets deciding whether to drop each
until it actually forwards one. Thus, every call to dualq_dequeue()
returns one packet if at least one is present in either queue,
otherwise it returns NULL at line 14. (Note 3)
Within each queue, the decision whether to drop or mark is taken as The dequeue pseudocode (Figure 10) is repeatedly called whenever the
follows (to simplify the explanation, it is assumed that U=1): lower layer is ready to forward a packet. It schedules one packet
for dequeuing (or zero if the queue is empty) then returns control to
the caller, so that it does not block while that packet is being
forwarded. While making this dequeue decision, it also makes the
necessary AQM decisions on dropping or marking. The alternative of
applying the AQMs at enqueue would shift some processing from the
critical time when each packet is dequeued. However, it would also
add a whole queue of delay to the control signals, making the control
loop very sloppy.
L4S: If the test at line 2 determines there is an L4S packet to The code is written assuming the AQMs are applied on dequeue (Note
dequeue, the tests at lines 3a and 3c determine whether to mark 1). All the dequeue code is contained within a large while loop so
it. The first is a simple test of whether the L4S queue (lq.byt() that if it decides to drop a packet, it will continue until it
in bytes) is greater than a step threshold T in bytes (Note 4). selects a packet to schedule. If both queues are empty, the routine
The second test is similar to the random ECN marking in RED, but returns NULL at line 20. Line 3 of the dequeue pseudocode is where
with the following differences: i) the marking function does not the conditional priority scheduler chooses between the L4S queue (lq)
start with a plateau of zero marking until a minimum threshold, and the Classic queue (cq). The time-shifted FIFO scheduler is shown
rather the marking probability starts to increase as soon as the at lines 28-33, which would be suitable if simplicity is paramount
queue is positive; ii) marking depends on queuing time, not bytes, (see Note 2).
in order to scale for any link rate without being reconfigured;
iii) marking of the L4S queue does not depend on itself, it Within each queue, the decision whether to forward, drop or mark is
depends on the queuing time of the _other_ (Classic) queue, where taken as follows (to simplify the explanation, it is assumed that
cq.sec() is the queuing time of the packet at the head of the U=1):
Classic queue (zero if empty); iv) marking depends on the
instantaneous queuing time (of the other Classic queue), not a L4S: If the test at line 3 determines there is an L4S packet to
smoothed average; v) the queue is compared with the maximum of U dequeue, the tests at lines 5b and 5c determine whether to mark
it. The first is a simple test of whether the L4S queue delay
(lq.time()) is greater than a step threshold T (Note 3). The
second test is similar to the random ECN marking in RED, but with
the following differences: ii) marking depends on queuing time,
not bytes, in order to scale for any link rate without being
reconfigured; ii) marking of the L4S queue does not depend on
itself, it depends on the queuing time of the _other_ (Classic)
queue, where cq.time() is the queuing time of the packet at the
head of the Classic queue (zero if empty); iii) marking depends on
the instantaneous queuing time (of the other Classic queue), not a
smoothed average; iv) the queue is compared with the maximum of U
random numbers (but if U=1, this is the same as the single random random numbers (but if U=1, this is the same as the single random
number used in RED). number used in RED).
Specifically, in line 3a the marking probability p_L is set to the Specifically, in line 5a the coupled marking probability p_CL is
Classic queueing time qc.sec() in seconds divided by the L4S set to the excess of the Classic queueing delay qc.time() above
scaling parameter 2^S_L, which represents the queuing time (in the minimum queuing delay threshold (minTh) all divided by the L4S
seconds) at which marking probability would hit 100%. Then in line scaling parameter range_L. range_L represents the queuing delay
3d (if U=1) the result is compared with a uniformly distributed (in seconds) added to minTh at which marking probability would hit
random number between 0 and 1, which ensures that marking 100%. Then in line 5c (if U=1) the result is compared with a
probability will linearly increase with queueing time. The uniformly distributed random number between 0 and 1, which ensures
scaling parameter is expressed as a power of 2 so that division that marking probability will linearly increase with queueing
can be implemented as a right bit-shift (>>) in line 3 of the time.
integer variant of the pseudocode (Figure 10).
Classic: If the test at line 7 determines that there is at least one Classic: If the scheduler at line 3 chooses to dequeue a Classic
Classic packet to dequeue, the test at line 9b determines whether packet and jumps to line 7, the test at line 10b determines
to drop it. But before that, line 8b updates Q_C, which is an whether to drop or mark it. But before that, line 9a updates Q_C,
exponentially weighted moving average (Note 5) of the queuing time which is an exponentially weighted moving average (Note 4) of the
in the Classic queue, where pkt.sec() is the instantaneous queuing time in the Classic queue, where qc.time() is the current
queueing time of the current Classic packet and alpha is the EWMA instantaneous queueing time of the Classic queue and gamma is the
constant for the classic queue. In line 8a, alpha is represented EWMA constant (default 1/32, see line 12 of the initialization
as an integer power of 2, so that in line 8 of the integer code function).
the division needed to weight the moving average can be
implemented by a right bit-shift (>> f_C).
Lines 9a and 9b implement the drop function. In line 9a the Lines 10a and 10b implement the Classic AQM. In line 10a the
averaged queuing time Q_C is divided by the Classic scaling averaged queuing time Q_C is divided by the Classic scaling
parameter 2^S_C, in the same way that queuing time was scaled for parameter range_C, in the same way that queuing time was scaled
L4S marking. This scaled queuing time is given the variable name for L4S marking. This scaled queuing time will be squared to
sqrt_p_C because it will be squared to compute Classic drop compute Classic drop probability so, before it is squared, it is
probability, so before it is squared it is effectively the square effectively the square root of the drop probability, hence it is
root of the drop probability. The squaring is done by comparing given the variable name sqrt_p_C. The squaring is done by
it with the maximum out of two random numbers (assuming U=1). comparing it with the maximum out of two random numbers (assuming
Comparing it with the maximum out of two is the same as the U=1). Comparing it with the maximum out of two is the same as the
logical `AND' of two tests, which ensures drop probability rises logical `AND' of two tests, which ensures drop probability rises
with the square of queuing time (Note 6). Again, the scaling with the square of queuing time.
parameter is expressed as a power of 2 so that division can be
implemented as a right bit-shift in line 9 of the integer
pseudocode.
The marking/dropping functions in each queue (lines 3 & 9) are two The AQM functions in each queue (lines 5c & 10b) are two cases of a
cases of a new generalization of RED called Curvy RED, motivated as new generalization of RED called Curvy RED, motivated as follows.
follows. When we compared the performance of our AQM with fq_CoDel When the performance of this AQM was compared with fq_CoDel and PIE,
and PIE, we came to the conclusion that their goal of holding queuing their goal of holding queuing delay to a fixed target seemed
delay to a fixed target is misguided [CRED_Insights]. As the number misguided [CRED_Insights]. As the number of flows increases, if the
of flows increases, if the AQM does not allow TCP to increase queuing AQM does not allow TCP to increase queuing delay, it has to introduce
delay, it has to introduce abnormally high levels of loss. Then loss abnormally high levels of loss. Then loss rather than queuing
rather than queuing becomes the dominant cause of delay for short becomes the dominant cause of delay for short flows, due to timeouts
flows, due to timeouts and tail losses. and tail losses.
Curvy RED constrains delay with a softened target that allows some Curvy RED constrains delay with a softened target that allows some
increase in delay as load increases. This is achieved by increasing increase in delay as load increases. This is achieved by increasing
drop probability on a convex curve relative to queue growth (the drop probability on a convex curve relative to queue growth (the
square curve in the Classic queue, if U=1). Like RED, the curve hugs square curve in the Classic queue, if U=1). Like RED, the curve hugs
the zero axis while the queue is shallow. Then, as load increases, the zero axis while the queue is shallow. Then, as load increases,
it introduces a growing barrier to higher delay. But, unlike RED, it it introduces a growing barrier to higher delay. But, unlike RED, it
requires only one parameter, the scaling, not three. The diadvantage requires only two parameters, not three. The disadvantage of Curvy
of Curvy RED is that it is not adapted to a wide range of RTTs. RED is that it is not adapted to a wide range of RTTs. Curvy RED can
Curvy RED can be used as is when the RTT range to support is limited be used as is when the RTT range to be supported is limited,
otherwise an adaptation mechanism is required. otherwise an adaptation mechanism is required.
There follows a summary listing of the two parameters used for each From our limited experiments with Curvy RED so far, recommended
of the two queues: values of these parameters are: S_C = -1; g_C = 5; T = 5 * MTU at the
link rate (about 1ms at 60Mb/s) for the range of base RTTs typical on
Classic: the public Internet. [CRED_Insights] explains why these parameters
are applicable whatever rate link this AQM implementation is deployed
on and how the parameters would need to be adjusted for a scenario
with a different range of RTTs (e.g. a data centre). The setting of
k depends on policy (see Section 2.5 and Appendix C respectively for
its recommended setting and guidance on alternatives).
S_C : The scaling factor of the dropping function scales Classic There is also a cUrviness parameter, U, which is a small positive
queuing times in the range [0, 2^(S_C)] seconds into a dropping integer. It is likely to take the same hard-coded value for all
probability in the range [0,1]. To make division efficient, it implementations, once experiments have determined a good value. Only
is constrained to be an integer power of two; U=1 has been used in experiments so far, but results might be even
better with U=2 or higher.
f_C : To smooth the queuing time of the Classic queue and make Notes:
multiplication efficient, we use a negative integer power of
two for the dimensionless EWMA constant, which we define as
alpha = 2^(-f_C).
L4S : 1. The alternative of applying the AQMs at enqueue would shift some
processing from the critical time when each packet is dequeued.
However, it would also add a whole queue of delay to the control
signals, making the control loop sloppier (for a typical RTT it
would double the Classic queue's feedback delay). On a platform
where packet timestamping is feasible, e.g. Linux, it is also
easiest to apply the AQMs at dequeue because that is where
queuing time is also measured.
S_L (and k'): As for the Classic queue, the scaling factor of 2. WRR better isolates the L4S queue from large delay bursts in the
the L4S marking function scales Classic queueing times in the Classic queue, but it is slightly less simple than TS-FIFO. If
range [0, 2^(S_L)] seconds into a probability in the range WRR were used, a low default Classic weight (e.g. 1/16) would
[0,1]. Note that S_L = S_C + k', where k' is the coupling need to be configured in place of the time shift in line 5 of the
between the queues. So S_L and k' count as only one parameter; initialization function (Figure 9).
k' is related to k in Equation (1) (Section 2.1) by k=2^k',
where both k and k' are constants. Then implementations can
avoid costly division by shifting p_L by k' bits to the right.
T : The queue size in bytes at which step threshold marking 3. A step function is shown for simplicity. A ramp function (see
starts in the L4S queue. Figure 5 and the discussion around it in Appendix A.1) is
recommended, because it is more general than a step and has the
potential to enable L4S congestion controls to converge more
rapidly.
{ToDo: These are the raw parameters used within the algorithm. A 4. An EWMA is only one possible way to filter bursts; other more
configuration front-end could accept more meaningful parameters and adaptive smoothing methods could be valid and it might be
convert them into these raw parameters.} appropriate to decrease the EWMA faster than it increases, e.g.
by using the minimum of the smoothed and instantaneous queue
delays, min(Q_C, qc.time()).
From our experiments so far, recommended values for these parameters B.2. Efficient Implementation of Curvy RED
are: S_C = -1; f_C = 5; T = 5 * MTU for the range of base RTTs
typical on the public Internet. [CRED_Insights] explains why these
parameters are applicable whatever rate link this AQM implementation
is deployed on and how the parameters would need to be adjusted for a
scenario with a different range of RTTs (e.g. a data centre) {ToDo
incorporate a summary of that report into this draft}. The setting of
k depends on policy (see Section 2.5 and Appendix C respectively for
its recommended setting and guidance on alternatives).
There is also a cUrviness parameter, U, which is a small positive Although code optimization depends on the platform, the following
integer. It is likely to take the same hard-coded value for all notes explain where the design of Curvy RED was particularly
implementations, once experiments have determined a good value. We motivated by efficient implementation.
have solely used U=1 in our experiments so far, but results might be
even better with U=2 or higher.
Note that the dropping function at line 9 calls maxrand(2*U), which The Classic AQM at line 10b calls maxrand(2*U), which gives twice as
gives twice as much curviness as the call to maxrand(U) in the much curviness as the call to maxrand(U) in the marking function at
marking function at line 3. This is the trick that implements the line 5c. This is the trick that implements the square rule in
square rule in equation (1) (Section 2.1). This is based on the fact equation (1) (Section 2.1). This is based on the fact that, given a
that, given a number X from 1 to 6, the probability that two dice number X from 1 to 6, the probability that two dice throws will both
throws will both be less than X is the square of the probability that be less than X is the square of the probability that one throw will
one throw will be less than X. So, when U=1, the L4S marking be less than X. So, when U=1, the L4S marking function is linear and
function is linear and the Classic dropping function is squared. If the Classic dropping function is squared. If U=2, L4S would be a
U=2, L4S would be a square function and Classic would be quartic. square function and Classic would be quartic. And so on.
And so on.
The maxrand(u) function in lines 16-21 simply generates u random The maxrand(u) function in lines 16-21 simply generates u random
numbers and returns the maximum (Note 7). Typically, maxrand(u) numbers and returns the maximum. Typically, maxrand(u) could be run
could be run in parallel out of band. For instance, if U=1, the in parallel out of band. For instance, if U=1, the Classic queue
Classic queue would require the maximum of two random numbers. So, would require the maximum of two random numbers. So, instead of
instead of calling maxrand(2*U) in-band, the maximum of every pair of calling maxrand(2*U) in-band, the maximum of every pair of values
values from a pseudorandom number generator could be generated out- from a pseudorandom number generator could be generated out-of-band,
of-band, and held in a buffer ready for the Classic queue to consume. and held in a buffer ready for the Classic queue to consume.
1: dualq_dequeue(lq, cq) { % Couples L4S & Classic queues, lq & cq 1: cred_dequeue(lq, cq, pkt) { % Couples L4S & Classic queues
2: if ( lq.dequeue(pkt) ) { 2: while ( lq.len() + cq.len() > 0 ) {
3: if ((lq.byt() > T) || ((cq.ns() >> (S_L-2)) > maxrand(U))) 3: if ( scheduler() == lq ) {
4: mark(pkt) 4: lq.dequeue(pkt) % L4S scheduled
5: return(pkt) % return the packet and stop here 5: if ((lq.time() > T) OR (cq.ns() >> (S_L-2) > maxrand(U)))
6: } 6: mark(pkt)
7: while ( cq.dequeue(pkt) ) { 7: } else {
8: Q_C += (pkt.ns() - Q_C) >> f_C % Classic Q EWMA 8: cq.dequeue(pkt) % Classic scheduled
9: if ( (Q_C >> (S_C-2) ) > maxrand(2*U) ) 9: Q_C += (cq.ns() - Q_C) >> g_C % Classic Q EWMA
10: drop(pkt) % Squared drop, redo loop 10: if ( (Q_C >> (S_C-2) ) > maxrand(2*U) ) {
11: else 11: if ( (ecn(pkt) == 0) { % ECN field = not-ECT
12: return(pkt) % return the packet and stop here 12: drop(pkt) % Squared drop, redo loop
13: } 13: continue % continue to the top of the while loop
14: return(NULL) % no packet to dequeue 14: }
15: } 15: mark(pkt)
16: }
17: }
18: return(pkt) % return the packet and stop here
19: }
20: return(NULL) % no packet to dequeue
21: }
Figure 10: Optimised Example Dequeue Pseudocode for Coupled DualQ AQM Figure 11: Optimised Example Dequeue Pseudocode for Coupled DualQ AQM
using Integer Arithmetic using Integer Arithmetic
Notes: The two ranges, range_L and range_C are expressed as powers of 2 so
that division can be implemented as a right bit-shift (>>) in lines 5
1. The drain rate of the queue can vary if it is scheduled relative and 10 of the integer variant of the pseudocode (Figure 11).
to other queues, or to cater for fluctuations in a wireless
medium. To auto-adjust to changes in drain rate, the queue must
be measured in time, not bytes or packets [CoDel]. In our Linux
implementation, it was easiest to measure queuing time at
dequeue. Queuing time can be estimated when a packet is enqueued
by measuring the queue length in bytes and dividing by the recent
drain rate.
2. An implementation has to use priority queueing, but it need not
implement strict priority.
3. If packets can be enqueued while processing dequeue code, an
implementer might prefer to place the while loop around both
queues so that it goes back to test again whether any L4S packets
arrived while it was dropping a Classic packet.
4. In order not to change too many factors at once, for now, we keep
the marking function for DCTCP-only traffic as similar as
possible to DCTCP. However, unlike DCTCP, all processing is at
dequeue, so we determine whether to mark a packet at the head of
the queue by the byte-length of the queue _behind_ it. We plan
to test whether using queuing time will work in all
circumstances, and if we find that the step can cause
oscillations, we will investigate replacing it with a steep
random marking curve.
5. An EWMA is only one possible way to filter bursts; other more For the integer variant of the pseudocode, an integer version of the
adaptive smoothing methods could be valid and it might be rand() function used at line 25 of the maxrand(function) in Figure 10
appropriate to decrease the EWMA faster than it increases. would be arranged to return an integer in the range 0 <= maxrand() <
2^32 (not shown). This would scale up all the floating point
probabilities in the range [0,1] by 2^32.
6. In practice at line 10 the Classic queue would probably test for Queuing delays are also scaled up by 2^32, but in two stages: i) In
ECN capability on the packet to determine whether to drop or mark lines 5 and 10 queuing times cq.ns() and pkt.ns() are returned in
the packet. However, for brevity such detail is omitted. All integer nanoseconds, making the values about 2^30 times larger than
packets classified into the L4S queue have to be ECN-capable, so when the units were seconds, ii) then in lines 3 and 9 an adjustment
no dropping logic is necessary at line 3. Nonetheless, L4S of -2 to the right bit-shift multiplies the result by 2^2, to
packets could be dropped by overload code (see Section 4.1). complete the scaling by 2^32.
7. In the integer variant of the pseudocode (Figure 10) real numbers In line 8 of the initialization function, the EWMA constant gamma is
are all represented as integers scaled up by 2^32. In lines 3 & represented as an integer power of 2, g_C, so that in line 9 of the
9 the function maxrand() is arranged to return an integer in the integer code the division needed to weight the moving average can be
range 0 <= maxrand() < 2^32. Queuing times are also scaled up by implemented by a right bit-shift (>> g_C).
2^32, but in two stages: i) In lines 3 and 8 queuing times
cq.ns() and pkt.ns() are returned in integer nanoseconds, making
the values about 2^30 times larger than when the units were
seconds, ii) then in lines 3 and 9 an adjustment of -2 to the
right bit-shift multiplies the result by 2^2, to complete the
scaling by 2^32.
Appendix C. Guidance on Controlling Throughput Equivalence Appendix C. Guidance on Controlling Throughput Equivalence
+---------------+------+-------+ +---------------+------+-------+
| RTT_C / RTT_L | Reno | Cubic | | RTT_C / RTT_L | Reno | Cubic |
+---------------+------+-------+ +---------------+------+-------+
| 1 | k'=1 | k'=0 | | 1 | k'=1 | k'=0 |
| 2 | k'=2 | k'=1 | | 2 | k'=2 | k'=1 |
| 3 | k'=2 | k'=2 | | 3 | k'=2 | k'=2 |
| 4 | k'=3 | k'=2 | | 4 | k'=3 | k'=2 |
skipping to change at page 40, line 32 skipping to change at page 49, line 11
for k', if it wants to slow DCTCP down to roughly the same throughput for k', if it wants to slow DCTCP down to roughly the same throughput
as Classic flows, to compensate for Classic flows slowing themselves as Classic flows, to compensate for Classic flows slowing themselves
down by causing themselves extra queuing delay. down by causing themselves extra queuing delay.
The values for k' in the table are derived from the formulae, which The values for k' in the table are derived from the formulae, which
was developed in [DCttH15]: was developed in [DCttH15]:
2^k' = 1.64 (RTT_reno / RTT_dc) (2) 2^k' = 1.64 (RTT_reno / RTT_dc) (2)
2^k' = 1.19 (RTT_cubic / RTT_dc ) (3) 2^k' = 1.19 (RTT_cubic / RTT_dc ) (3)
For localized traffic from a particular ISP's data centre, we used For localized traffic from a particular ISP's data centre, using the
the measured RTTs to calculate that a value of k'=3 (equivalant to measured RTTs, it was calculated that a value of k'=3 (equivalant to
k=8) would achieve throughput equivalence, and our experiments k=8) would achieve throughput equivalence, and experiments verified
verified the formula very closely. the formula very closely.
For a typical mix of RTTs from local data centres and across the For a typical mix of RTTs from local data centres and across the
general Internet, a value of k'=1 (equivalent to k=2) is recommended general Internet, a value of k'=1 (equivalent to k=2) is recommended
as a good workable compromise. as a good workable compromise.
Appendix D. Open Issues
Most of the following open issues are also tagged '{ToDo}' at the
appropriate point in the document:
PI2 appendix: scaling of alpha & beta, esp. dependence of beta_U
on Tupdate
Curvy RED appendix: complete the unfinished parts
Authors' Addresses Authors' Addresses
Koen De Schepper Koen De Schepper
Nokia Bell Labs Nokia Bell Labs
Antwerp Antwerp
Belgium Belgium
Email: koen.de_schepper@nokia.com Email: koen.de_schepper@nokia.com
URI: https://www.bell-labs.com/usr/koen.de_schepper URI: https://www.bell-labs.com/usr/koen.de_schepper
Bob Briscoe (editor) Bob Briscoe (editor)
CableLabs CableLabs
UK UK
Email: ietf@bobbriscoe.net Email: ietf@bobbriscoe.net
URI: http://bobbriscoe.net/ URI: http://bobbriscoe.net/
Olga Bondarenko Greg White
Simula Research Lab CableLabs
Lysaker Louisville, CO
Norway US
Email: olgabnd@gmail.com
URI: https://www.simula.no/people/olgabo
Ing-jyh Tsang
Nokia
Antwerp
Belgium
Email: ing-jyh.tsang@nokia.com Email: G.White@CableLabs.com
 End of changes. 189 change blocks. 
844 lines changed or deleted 1196 lines changed or added

This html diff was produced by rfcdiff 1.47. The latest version is available from http://tools.ietf.org/tools/rfcdiff/