draft-ietf-aqm-recommendation-11.txt   rfc7567.txt 
Network Working Group F. Baker, Ed. Internet Engineering Task Force (IETF) F. Baker, Ed.
Internet-Draft Cisco Systems Request for Comments: 7567 Cisco Systems
Obsoletes: 2309 (if approved) G. Fairhurst, Ed. BCP: 197 G. Fairhurst, Ed.
Intended status: Best Current Practice University of Aberdeen Obsoletes: 2309 University of Aberdeen
Expires: August 29, 2015 February 25, 2015 Category: Best Current Practice July 2015
ISSN: 2070-1721
IETF Recommendations Regarding Active Queue Management IETF Recommendations Regarding Active Queue Management
draft-ietf-aqm-recommendation-11
Abstract Abstract
This memo presents recommendations to the Internet community This memo presents recommendations to the Internet community
concerning measures to improve and preserve Internet performance. It concerning measures to improve and preserve Internet performance. It
presents a strong recommendation for testing, standardization, and presents a strong recommendation for testing, standardization, and
widespread deployment of active queue management (AQM) in network widespread deployment of active queue management (AQM) in network
devices, to improve the performance of today's Internet. It also devices to improve the performance of today's Internet. It also
urges a concerted effort of research, measurement, and ultimate urges a concerted effort of research, measurement, and ultimate
deployment of AQM mechanisms to protect the Internet from flows that deployment of AQM mechanisms to protect the Internet from flows that
are not sufficiently responsive to congestion notification. are not sufficiently responsive to congestion notification.
The note replaces the recommendations of RFC 2309 based on fifteen Based on 15 years of experience and new research, this document
years of experience and new research. replaces the recommendations of RFC 2309.
Status of This Memo Status of This Memo
This Internet-Draft is submitted in full conformance with the This memo documents an Internet Best Current Practice.
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months This document is a product of the Internet Engineering Task Force
and may be updated, replaced, or obsoleted by other documents at any (IETF). It represents the consensus of the IETF community. It has
time. It is inappropriate to use Internet-Drafts as reference received public review and has been approved for publication by the
material or to cite them other than as "work in progress." Internet Engineering Steering Group (IESG). Further information on
BCPs is available in Section 2 of RFC 5741.
This Internet-Draft will expire on August 29, 2015. Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc7567.
Copyright Notice Copyright Notice
Copyright (c) 2015 IETF Trust and the persons identified as the Copyright (c) 2015 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
(http://trustee.ietf.org/license-info) in effect on the date of (http://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
skipping to change at page 2, line 26 skipping to change at page 3, line 7
modifications of such material outside the IETF Standards Process. modifications of such material outside the IETF Standards Process.
Without obtaining an adequate license from the person(s) controlling Without obtaining an adequate license from the person(s) controlling
the copyright in such materials, this document may not be modified the copyright in such materials, this document may not be modified
outside the IETF Standards Process, and derivative works of it may outside the IETF Standards Process, and derivative works of it may
not be created outside the IETF Standards Process, except to format not be created outside the IETF Standards Process, except to format
it for publication as an RFC or to translate it into languages other it for publication as an RFC or to translate it into languages other
than English. than English.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1. Congestion Collapse . . . . . . . . . . . . . . . . . . . 3 1.1. Congestion Collapse . . . . . . . . . . . . . . . . . . . 4
1.2. Active Queue Management to Manage Latency . . . . . . . . 4 1.2. Active Queue Management to Manage Latency . . . . . . . . 5
1.3. Document Overview . . . . . . . . . . . . . . . . . . . . 5 1.3. Document Overview . . . . . . . . . . . . . . . . . . . . 6
1.4. Changes to the recommendations of RFC2309 . . . . . . . . 6 1.4. Changes to the Recommendations of RFC 2309 . . . . . . . 7
1.5. Requirements Language . . . . . . . . . . . . . . . . . . 6 1.5. Requirements Language . . . . . . . . . . . . . . . . . . 7
2. The Need For Active Queue Management . . . . . . . . . . . . 6 2. The Need for Active Queue Management . . . . . . . . . . . . 7
2.1. AQM and Multiple Queues . . . . . . . . . . . . . . . . . 10 2.1. AQM and Multiple Queues . . . . . . . . . . . . . . . . . 11
2.2. AQM and Explicit Congestion Marking (ECN) . . . . . . . . 10 2.2. AQM and Explicit Congestion Marking (ECN) . . . . . . . . 12
2.3. AQM and Buffer Size . . . . . . . . . . . . . . . . . . . 11 2.3. AQM and Buffer Size . . . . . . . . . . . . . . . . . . . 12
3. Managing Aggressive Flows . . . . . . . . . . . . . . . . . . 11 3. Managing Aggressive Flows . . . . . . . . . . . . . . . . . . 13
4. Conclusions and Recommendations . . . . . . . . . . . . . . . 14 4. Conclusions and Recommendations . . . . . . . . . . . . . . . 16
4.1. Operational deployments SHOULD use AQM procedures . . . . 15 4.1. Operational Deployments SHOULD Use AQM Procedures . . . . 17
4.2. Signaling to the transport endpoints . . . . . . . . . . 16 4.2. Signaling to the Transport Endpoints . . . . . . . . . . 17
4.2.1. AQM and ECN . . . . . . . . . . . . . . . . . . . . . 17 4.2.1. AQM and ECN . . . . . . . . . . . . . . . . . . . . . 18
4.3. AQM algorithm deployment SHOULD NOT require operational 4.3. AQM Algorithm Deployment SHOULD NOT Require Operational
tuning . . . . . . . . . . . . . . . . . . . . . . . . . 18 Tuning . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.4. AQM algorithms SHOULD respond to measured congestion, not 4.4. AQM Algorithms SHOULD Respond to Measured Congestion, Not
application profiles. . . . . . . . . . . . . . . . . . . 19 Application Profiles . . . . . . . . . . . . . . . . . . 21
4.5. AQM algorithms SHOULD NOT be dependent on specific 4.5. AQM Algorithms SHOULD NOT Be Dependent on Specific
transport protocol behaviours . . . . . . . . . . . . . . 20 Transport Protocol Behaviors . . . . . . . . . . . . . . 22
4.6. Interactions with congestion control algorithms . . . . . 21 4.6. Interactions with Congestion Control Algorithms . . . . . 22
4.7. The need for further research . . . . . . . . . . . . . . 22 4.7. The Need for Further Research . . . . . . . . . . . . . . 23
5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 23 5. Security Considerations . . . . . . . . . . . . . . . . . . . 25
6. Security Considerations . . . . . . . . . . . . . . . . . . . 23 6. Privacy Considerations . . . . . . . . . . . . . . . . . . . 25
7. Privacy Considerations . . . . . . . . . . . . . . . . . . . 23 7. References . . . . . . . . . . . . . . . . . . . . . . . . . 25
8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 24 7.1. Normative References . . . . . . . . . . . . . . . . . . 25
9. References . . . . . . . . . . . . . . . . . . . . . . . . . 24 7.2. Informative References . . . . . . . . . . . . . . . . . 26
9.1. Normative References . . . . . . . . . . . . . . . . . . 24 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . 31
9.2. Informative References . . . . . . . . . . . . . . . . . 25 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 31
Appendix A. Change Log . . . . . . . . . . . . . . . . . . . . . 28
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 30
1. Introduction 1. Introduction
The Internet protocol architecture is based on a connectionless end- The Internet protocol architecture is based on a connectionless end-
to-end packet service using the Internet Protocol, whether IPv4 to-end packet service using the Internet Protocol, whether IPv4
[RFC0791] or IPv6 [RFC2460]. The advantages of its connectionless [RFC791] or IPv6 [RFC2460]. The advantages of its connectionless
design: flexibility and robustness, have been amply demonstrated. design -- flexibility and robustness -- have been amply demonstrated.
However, these advantages are not without cost: careful design is However, these advantages are not without cost: careful design is
required to provide good service under heavy load. In fact, lack of required to provide good service under heavy load. In fact, lack of
attention to the dynamics of packet forwarding can result in severe attention to the dynamics of packet forwarding can result in severe
service degradation or "Internet meltdown". This phenomenon was service degradation or "Internet meltdown". This phenomenon was
first observed during the early growth phase of the Internet in the first observed during the early growth phase of the Internet in the
mid 1980s [RFC0896][RFC0970], and is technically called "congestion mid 1980s [RFC896] [RFC970]; it is technically called "congestion
collapse" and was a key focus of RFC2309. collapse" and was a key focus of RFC 2309.
Although wide-scale congestion collapse is not common in the Although wide-scale congestion collapse is not common in the
Internet, the presence of localised congestion collapse is by no Internet, the presence of localized congestion collapse is by no
means rare. It is therefore important to continue to avoid means rare. It is therefore important to continue to avoid
congestion collapse. congestion collapse.
Since 1998, when RFC2309 was written, the Internet has become used Since 1998, when RFC 2309 was written, the Internet has become used
for a variety of traffic. In the current Internet, low latency is for a variety of traffic. In the current Internet, low latency is
extremely important for many interactive and transaction-based extremely important for many interactive and transaction-based
applications. The same type of technology that RFC2309 advocated for applications. The same type of technology that RFC 2309 advocated
combating congestion collapse is also effective at limiting delays to for combating congestion collapse is also effective at limiting
reduce the interaction delay (latency) experienced by applications delays to reduce the interaction delay (latency) experienced by
[Bri15]. High or unpredictable latency can impact the performance of applications [Bri15]. High or unpredictable latency can impact the
the control loops used by ene-to-end protocols (including congestion performance of the control loops used by end-to-end protocols
control algorithms using TCP). There is now also a focus on reducing (including congestion control algorithms using TCP). There is now
network latency using the same technology. also a focus on reducing network latency using the same technology.
The mechanisms decsribed in this document may be implemented in The mechanisms described in this document may be implemented in
network devices on the path between end-points that include routers, network devices on the path between endpoints that include routers,
switches, and other network middleboxes. The methods may also be switches, and other network middleboxes. The methods may also be
implemented in the networking stacks within endpoint devices that implemented in the networking stacks within endpoint devices that
connect to the network. connect to the network.
1.1. Congestion Collapse 1.1. Congestion Collapse
The original fix for Internet meltdown was provided by Van Jacobsen. The original fix for Internet meltdown was provided by Van Jacobsen.
Beginning in 1986, Jacobsen developed the congestion avoidance Beginning in 1986, Jacobsen developed the congestion avoidance
mechanisms [Jacobson88] that are now required for implementations of mechanisms [Jacobson88] that are now required for implementations of
the Transport Control Protocol (TCP) [RFC0793] [RFC1122]. ([RFC7414] the Transport Control Protocol (TCP) [RFC793] [RFC1122]. ([RFC7414]
provides a roadmap to help identify TCP-related documents.) These provides a roadmap to help identify TCP-related documents.) These
mechanisms operate in Internet hosts to cause TCP connections to mechanisms operate in Internet hosts to cause TCP connections to
"back off" during congestion. We say that TCP flows are "responsive" "back off" during congestion. We say that TCP flows are "responsive"
to congestion signals (i.e., packets that are dropped or marked with to congestion signals (i.e., packets that are dropped or marked with
explicit congestion notification [RFC3168]). It is primarily these explicit congestion notification [RFC3168]). It is primarily these
TCP congestion avoidance algorithms that prevent the congestion TCP congestion avoidance algorithms that prevent the congestion
collapse of today's Internet. Similar algorithms are specified for collapse of today's Internet. Similar algorithms are specified for
other non-TCP transports. other non-TCP transports.
However, that is not the end of the story. Considerable research has However, that is not the end of the story. Considerable research has
skipping to change at page 4, line 28 skipping to change at page 5, line 22
provide good service in all circumstances. Basically, there is a provide good service in all circumstances. Basically, there is a
limit to how much control can be accomplished from the edges of the limit to how much control can be accomplished from the edges of the
network. Some mechanisms are needed in network devices to complement network. Some mechanisms are needed in network devices to complement
the endpoint congestion avoidance mechanisms. These mechanisms may the endpoint congestion avoidance mechanisms. These mechanisms may
be implemented in network devices. be implemented in network devices.
1.2. Active Queue Management to Manage Latency 1.2. Active Queue Management to Manage Latency
Internet latency has become a focus of attention to increase the Internet latency has become a focus of attention to increase the
responsiveness of Internet applications and protocols. One major responsiveness of Internet applications and protocols. One major
source of delay is the build-up of queues in network devices. source of delay is the buildup of queues in network devices.
Queueing occurs whenever the arrival rate of data at the ingress to a Queueing occurs whenever the arrival rate of data at the ingress to a
device exceeds the current egress rate. Such queueing is normal in a device exceeds the current egress rate. Such queueing is normal in a
packet-switched network and is often necessary to absorb bursts in packet-switched network and is often necessary to absorb bursts in
transmission and perform statistical multiplexing of traffic, but transmission and perform statistical multiplexing of traffic, but
excessive queueing can lead to unwanted delay, reducing the excessive queueing can lead to unwanted delay, reducing the
performance of some Internet applications. performance of some Internet applications.
RFC 2309 introduced the concept of "Active Queue Management" (AQM), a RFC 2309 introduced the concept of "Active Queue Management" (AQM), a
class of technologies that, by signaling to common congestion- class of technologies that, by signaling to common congestion-
controlled transports such as TCP, manages the size of queues that controlled transports such as TCP, manages the size of queues that
build in network buffers. RFC 2309 also describes a specific AQM build in network buffers. RFC 2309 also describes a specific AQM
algorithm, Random Early Detection (RED), and recommends that this be algorithm, Random Early Detection (RED), and recommends that this be
widely implemented and used by default in routers. widely implemented and used by default in routers.
With an appropriate set of parameters, RED is an effective algorithm. With an appropriate set of parameters, RED is an effective algorithm.
However, dynamically predicting this set of parameters was found to However, dynamically predicting this set of parameters was found to
be difficult. As a result, RED has not been enabled by default, and be difficult. As a result, RED has not been enabled by default, and
its present use in the Internet is limited. Other AQM algorithms its present use in the Internet is limited. Other AQM algorithms
have been developed since RC2309 was published, some of which are have been developed since RFC 2309 was published, some of which are
self-tuning within a range of applicability. Hence, while this memo self-tuning within a range of applicability. Hence, while this memo
continues to recommend the deployment of AQM, it no longer recommends continues to recommend the deployment of AQM, it no longer recommends
that RED or any other specific algorithm is used as a default; that RED or any other specific algorithm is used by default. It
instead it provides recommendations on how to select appropriate instead provides recommendations on IETF processes for the selection
algorithms and that a recommended algorithm is able to automate any of appropriate algorithms, and especially that a recommended
required tuning for common deployment scenarios. algorithm is able to automate any required tuning for common
deployment scenarios.
Deploying AQM in the network can significantly reduce the latency Deploying AQM in the network can significantly reduce the latency
across an Internet path and since writing RFC2309, this has become a across an Internet path, and, since the writing of RFC 2309, this has
key motivation for using AQM in the Internet. In the context of AQM, become a key motivation for using AQM in the Internet. In the
it is useful to distinguish between two related classes of context of AQM, it is useful to distinguish between two related
algorithms: "queue management" versus "scheduling" algorithms. To a classes of algorithms: "queue management" versus "scheduling"
rough approximation, queue management algorithms manage the length of algorithms. To a rough approximation, queue management algorithms
packet queues by marking or dropping packets when necessary or manage the length of packet queues by marking or dropping packets
appropriate, while scheduling algorithms determine which packet to when necessary or appropriate, while scheduling algorithms determine
send next and are used primarily to manage the allocation of which packet to send next and are used primarily to manage the
bandwidth among flows. While these two mechanisms are closely allocation of bandwidth among flows. While these two mechanisms are
related, they address different performance issues and operate on closely related, they address different performance issues and
different timescales. Both may be used in combination. operate on different timescales. Both may be used in combination.
1.3. Document Overview 1.3. Document Overview
The discussion in this memo applies to "best-effort" traffic, which The discussion in this memo applies to "best-effort" traffic, which
is to say, traffic generated by applications that accept the is to say, traffic generated by applications that accept the
occasional loss, duplication, or reordering of traffic in flight. It occasional loss, duplication, or reordering of traffic in flight. It
also applies to other traffic, such as real-time traffic that can also applies to other traffic, such as real-time traffic that can
adapt its sending rate to reduce loss and/or delay. It is most adapt its sending rate to reduce loss and/or delay. It is most
effective when the adaption occurs on time scales of a single Round effective when the adaption occurs on timescales of a single Round-
Trip Time (RTT) or a small number of RTTs, for elastic traffic Trip Time (RTT) or a small number of RTTs, for elastic traffic
[RFC1633]. [RFC1633].
Two performance issues are highlighted: Two performance issues are highlighted:
The first issue is the need for an advanced form of queue management The first issue is the need for an advanced form of queue management
that we call "Active Queue Management", AQM. Section 2 summarizes that we call "Active Queue Management", AQM. Section 2 summarizes
the benefits that active queue management can bring. A number of AQM the benefits that active queue management can bring. A number of AQM
procedures are described in the literature, with different procedures are described in the literature, with different
characteristics. This document does not recommend any of them in characteristics. This document does not recommend any of them in
particular, but does make recommendations that ideally would affect particular, but it does make recommendations that ideally would
the choice of procedure used in a given implementation. affect the choice of procedure used in a given implementation.
The second issue, discussed in Section 4 of this memo, is the The second issue, discussed in Section 4 of this memo, is the
potential for future congestion collapse of the Internet due to flows potential for future congestion collapse of the Internet due to flows
that are unresponsive, or not sufficiently responsive, to congestion that are unresponsive, or not sufficiently responsive, to congestion
indications. Unfortunately, while scheduling can mitigate some of indications. Unfortunately, while scheduling can mitigate some of
the side-effects of sharing a network queue with an unresponsive the side effects of sharing a network queue with an unresponsive
flow, there is currently no consensus solution to controlling the flow, there is currently no consensus solution to controlling the
congestion caused by such aggressive flows. Methods such as congestion caused by such aggressive flows. Methods such as
congestion exposure (ConEx) [RFC6789] offer a framework [CONEX] that congestion exposure (ConEx) [RFC6789] offer a framework [CONEX] that
can update network devices to alleviate these effects. Significant can update network devices to alleviate these effects. Significant
research and engineering will be required before any solution will be research and engineering will be required before any solution will be
available. It is imperative that work to mitigate the impact of available. It is imperative that work to mitigate the impact of
unresponsive flows is energetically pursued, to ensure acceptable unresponsive flows is energetically pursued to ensure acceptable
performance and the future stability of the Internet. performance and the future stability of the Internet.
Section 4 concludes the memo with a set of recommendations to the Section 4 concludes the memo with a set of recommendations to the
Internet community on the use of AQM and recommendations for defining Internet community on the use of AQM and recommendations for defining
AQM algorithms. AQM algorithms.
1.4. Changes to the recommendations of RFC2309 1.4. Changes to the Recommendations of RFC 2309
This memo replaces the recommendations in [RFC2309], which resulted This memo replaces the recommendations in [RFC2309], which resulted
from past discussions of end-to-end performance, Internet congestion, from past discussions of end-to-end performance, Internet congestion,
and RED in the End-to-End Research Group of the Internet Research and RED in the End-to-End Research Group of the Internet Research
Task Force (IRTF). It follows experience with this and other Task Force (IRTF). It results from experience with RED and other
algorithms, and the AQM discussion within the IETF [AQM-WG]. algorithms, and the AQM discussion within the IETF [AQM-WG].
While RFC2309 described AQM in terms of the length of a queue. This Whereas RFC 2309 described AQM in terms of the length of a queue,
memo changes this, to use AQM to refer to any method that allows this memo uses AQM to refer to any method that allows network devices
network devices to control either the queue length and/or the mean to control the queue length and/or the mean time that a packet spends
time that a packet spends in a queue. in a queue.
This memo also explicitly obsoletes the recommendation that Random This memo also explicitly obsoletes the recommendation that Random
Early Detection (RED) was to be used as the default AQM mechanism for Early Detection (RED) be used as the default AQM mechanism for the
the Internet. This is replaced by a detailed set of recommendations Internet. This is replaced by a detailed set of recommendations for
for selecting an appropriate AQM algorithm. As in RFC2309, this memo selecting an appropriate AQM algorithm. As in RFC 2309, this memo
also motivates the need for continued research, but clarifies the illustrates the need for continued research. It also clarifies the
research with examples appropriate at the time that this memo is research needed with examples appropriate at the time that this memo
published. is published.
1.5. Requirements Language 1.5. Requirements Language
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]. document are to be interpreted as described in [RFC2119].
2. The Need For Active Queue Management 2. The Need for Active Queue Management
Active Queue Management (AQM) is a method that allows network devices Active Queue Management (AQM) is a method that allows network devices
to control the queue length or the mean time that a packet spends in to control the queue length or the mean time that a packet spends in
a queue. Although AQM can be applied across a range of deployment a queue. Although AQM can be applied across a range of deployment
environments, the recommendations in this document are directed to environments, the recommendations in this document are for use in the
use in the general Internet. It is expected that the principles and general Internet. It is expected that the principles and guidance
guidance are also applicable to a wide range of environments, but may are also applicable to a wide range of environments, but they may
require tuning for specific types of link/network (e.g. to require tuning for specific types of links or networks (e.g., to
accommodate the traffic patterns found in data centres, the accommodate the traffic patterns found in data centers, the
challenges of wireless infrastructure, or the higher delay challenges of wireless infrastructure, or the higher delay
encountered on satellite Internet links). The remainder of this encountered on satellite Internet links). The remainder of this
section identifies the need for AQM and the advantages of deploying section identifies the need for AQM and the advantages of deploying
AQM methods. AQM methods.
The traditional technique for managing the queue length in a network The traditional technique for managing the queue length in a network
device is to set a maximum length (in terms of packets) for each device is to set a maximum length (in terms of packets) for each
queue, accept packets for the queue until the maximum length is queue, accept packets for the queue until the maximum length is
reached, then reject (drop) subsequent incoming packets until the reached, then reject (drop) subsequent incoming packets until the
queue decreases because a packet from the queue has been transmitted. queue decreases because a packet from the queue has been transmitted.
This technique is known as "tail drop", since the packet that arrived This technique is known as "tail drop", since the packet that arrived
most recently (i.e., the one on the tail of the queue) is dropped most recently (i.e., the one on the tail of the queue) is dropped
when the queue is full. This method has served the Internet well for when the queue is full. This method has served the Internet well for
years, but it has four important drawbacks: years, but it has four important drawbacks:
1. Full Queues 1. Full Queues
The tail drop discipline allows queues to maintain a full (or, The "tail drop" discipline allows queues to maintain a full (or,
almost full) status for long periods of time, since tail drop almost full) status for long periods of time, since tail drop
signals congestion (via a packet drop) only when the queue has signals congestion (via a packet drop) only when the queue has
become full. It is important to reduce the steady-state queue become full. It is important to reduce the steady-state queue
size, and this is perhaps the most important goal for queue size, and this is perhaps the most important goal for queue
management. management.
The naive assumption might be that there is a simple tradeoff The naive assumption might be that there is a simple trade-off
between delay and throughput, and that the recommendation that between delay and throughput, and that the recommendation that
queues be maintained in a "non-full" state essentially translates queues be maintained in a "non-full" state essentially translates
to a recommendation that low end-to-end delay is more important to a recommendation that low end-to-end delay is more important
than high throughput. However, this does not take into account than high throughput. However, this does not take into account
the critical role that packet bursts play in Internet the critical role that packet bursts play in Internet
performance. For example, even though TCP constrains the performance. For example, even though TCP constrains the
congestion window of a flow, packets often arrive at network congestion window of a flow, packets often arrive at network
devices in bursts [Leland94]. If the queue is full or almost devices in bursts [Leland94]. If the queue is full or almost
full, an arriving burst will cause multiple packets to be dropped full, an arriving burst will cause multiple packets to be dropped
from the same flow. Bursts of loss can result in a global from the same flow. Bursts of loss can result in a global
synchronization of flows throttling back, followed by a sustained synchronization of flows throttling back, followed by a sustained
period of lowered link utilization, reducing overall throughput period of lowered link utilization, reducing overall throughput
[Flo94], [Zha90] [Flo94] [Zha90].
The goal of buffering in the network is to absorb data bursts and The goal of buffering in the network is to absorb data bursts and
to transmit them during the (hopefully) ensuing bursts of to transmit them during the (hopefully) ensuing bursts of
silence. This is essential to permit transmission of bursts of silence. This is essential to permit transmission of bursts of
data. Normally small queues are preferred in network devices, data. Queues that are normally small are preferred in network
with sufficient queue capacity to absorb the bursts. The devices, with sufficient queue capacity to absorb the bursts.
counter-intuitive result is that maintaining normally-small The counterintuitive result is that maintaining queues that are
queues can result in higher throughput as well as lower end-to- normally small can result in higher throughput as well as lower
end delay. In summary, queue limits should not reflect the end-to-end delay. In summary, queue limits should not reflect
steady state queues we want to be maintained in the network; the steady-state queues we want to be maintained in the network;
instead, they should reflect the size of bursts that a network instead, they should reflect the size of bursts that a network
device needs to absorb. device needs to absorb.
2. Lock-Out 2. Lock-Out
In some situations tail drop allows a single connection or a few In some situations tail drop allows a single connection or a few
flows to monopolize the queue space starving other connections, flows to monopolize the queue space, thereby starving other
preventing them from getting room in the queue [Flo92]. connections, preventing them from getting room in the queue
[Flo92].
3. Mitigating the Impact of Packet Bursts 3. Mitigating the Impact of Packet Bursts
Large burst of packets can delay other packets, disrupting the A large burst of packets can delay other packets, disrupting the
control loop (e.g. the pacing of flows by the TCP ACK-Clock), and control loop (e.g., the pacing of flows by the TCP ACK clock),
reducing the performance of flows that share a common bottleneck. and reducing the performance of flows that share a common
bottleneck.
4. Control loop synchronization 4. Control Loop Synchronization
Congestion control, like other end-to-end mechanisms, introduces Congestion control, like other end-to-end mechanisms, introduces
a control loop between hosts. Sessions that share a common a control loop between hosts. Sessions that share a common
network bottleneck can therefore become synchronised, introducing network bottleneck can therefore become synchronized, introducing
periodic disruption (e.g. jitter/loss). "lock-out" is often also periodic disruption (e.g., jitter/loss). "Lock-out" is often
the result of synchronization or other timing effects also the result of synchronization or other timing effects
Besides tail drop, two alternative queue management disciplines that Besides tail drop, two alternative queue management disciplines that
can be applied when a queue becomes full are "random drop on full" or can be applied when a queue becomes full are "random drop on full" or
"head drop on full". When a new packet arrives at a full queue using "head drop on full". When a new packet arrives at a full queue using
the random drop on full discipline, the network device drops a the "random drop on full" discipline, the network device drops a
randomly selected packet from the queue (which can be an expensive randomly selected packet from the queue (this can be an expensive
operation, since it naively requires an O(N) walk through the packet operation, since it naively requires an O(N) walk through the packet
queue). When a new packet arrives at a full queue using the head queue). When a new packet arrives at a full queue using the "head
drop on full discipline, the network device drops the packet at the drop on full" discipline, the network device drops the packet at the
front of the queue [Lakshman96]. Both of these solve the lock-out front of the queue [Lakshman96]. Both of these solve the lock-out
problem, but neither solves the full-queues problem described above. problem, but neither solves the full-queues problem described above.
We know in general how to solve the full-queues problem for In general, we know how to solve the full-queues problem for
"responsive" flows, i.e., those flows that throttle back in response "responsive" flows, i.e., those flows that throttle back in response
to congestion notification. In the current Internet, dropped packets to congestion notification. In the current Internet, dropped packets
provide a critical mechanism indicating congestion notification to provide a critical mechanism indicating congestion notification to
hosts. The solution to the full-queues problem is for network hosts. The solution to the full-queues problem is for network
devices to drop or ECN-mark packets before a queue becomes full, so devices to drop or ECN-mark packets before a queue becomes full, so
that hosts can respond to congestion before buffers overflow. We that hosts can respond to congestion before buffers overflow. We
call such a proactive approach AQM. By dropping or ECN-marking call such a proactive approach AQM. By dropping or ECN-marking
packets before buffers overflow, AQM allows network devices to packets before buffers overflow, AQM allows network devices to
control when and how many packets to drop. control when and how many packets to drop.
In summary, an active queue management mechanism can provide the In summary, an active queue management mechanism can provide the
following advantages for responsive flows. following advantages for responsive flows.
1. Reduce number of packets dropped in network devices 1. Reduce number of packets dropped in network devices
Packet bursts are an unavoidable aspect of packet networks Packet bursts are an unavoidable aspect of packet networks
[Willinger95]. If all the queue space in a network device is [Willinger95]. If all the queue space in a network device is
already committed to "steady state" traffic or if the buffer already committed to "steady-state" traffic or if the buffer
space is inadequate, then the network device will have no ability space is inadequate, then the network device will have no ability
to buffer bursts. By keeping the average queue size small, AQM to buffer bursts. By keeping the average queue size small, AQM
will provide greater capacity to absorb naturally-occurring will provide greater capacity to absorb naturally occurring
bursts without dropping packets. bursts without dropping packets.
Furthermore, without AQM, more packets will be dropped when a Furthermore, without AQM, more packets will be dropped when a
queue does overflow. This is undesirable for several reasons. queue does overflow. This is undesirable for several reasons.
First, with a shared queue and the tail drop discipline, this can First, with a shared queue and the "tail drop" discipline, this
result in unnecessary global synchronization of flows, resulting can result in unnecessary global synchronization of flows,
in lowered average link utilization, and hence lowered network resulting in lowered average link utilization and, hence, lowered
throughput. Second, unnecessary packet drops represent a waste network throughput. Second, unnecessary packet drops represent a
of network capacity on the path before the drop point. waste of network capacity on the path before the drop point.
While AQM can manage queue lengths and reduce end-to-end latency While AQM can manage queue lengths and reduce end-to-end latency
even in the absence of end-to-end congestion control, it will be even in the absence of end-to-end congestion control, it will be
able to reduce packet drops only in an environment that continues able to reduce packet drops only in an environment that continues
to be dominated by end-to-end congestion control. to be dominated by end-to-end congestion control.
2. Provide a lower-delay interactive service 2. Provide a lower-delay interactive service
By keeping a small average queue size, AQM will reduce the delays By keeping a small average queue size, AQM will reduce the delays
experienced by flows. This is particularly important for experienced by flows. This is particularly important for
interactive applications such as short web transfers, POP/IMAP, interactive applications such as short web transfers, POP/IMAP,
DNS, terminal traffic (telnet, ssh, mosh, RDP, etc), gaming or DNS, terminal traffic (Telnet, SSH, Mosh, RDP, etc.), gaming or
interactive audio-video sessions, whose subjective (and interactive audio-video sessions, whose subjective (and
objective) performance is better when the end-to-end delay is objective) performance is better when the end-to-end delay is
low. low.
3. Avoid lock-out behavior 3. Avoid lock-out behavior
AQM can prevent lock-out behavior by ensuring that there will AQM can prevent lock-out behavior by ensuring that there will
almost always be a buffer available for an incoming packet. For almost always be a buffer available for an incoming packet. For
the same reason, AQM can prevent a bias against low capacity, but the same reason, AQM can prevent a bias against low-capacity, but
highly bursty, flows. highly bursty, flows.
Lock-out is undesirable because it constitutes a gross unfairness Lock-out is undesirable because it constitutes a gross unfairness
among groups of flows. However, we stop short of calling this among groups of flows. However, we stop short of calling this
benefit "increased fairness", because general fairness among benefit "increased fairness", because general fairness among
flows requires per-flow state, which is not provided by queue flows requires per-flow state, which is not provided by queue
management. For example, in a network device using AQM with only management. For example, in a network device using AQM with only
FIFO scheduling, two TCP flows may receive very different share FIFO scheduling, two TCP flows may receive very different shares
of the network capacity simply because they have different round- of the network capacity simply because they have different RTTs
trip times [Floyd91], and a flow that does not use congestion [Floyd91], and a flow that does not use congestion control may
control may receive more capacity than a flow that does. AQM can receive more capacity than a flow that does. AQM can therefore
therefore be combined with a scheduling mechanism that divides be combined with a scheduling mechanism that divides network
network traffic between multiple queues (section 2.1). traffic between multiple queues (Section 2.1).
4. Reduce the probability of control loop synchronization 4. Reduce the probability of control loop synchronization
The probability of network control loop synchronization can be The probability of network control loop synchronization can be
reduced if network devices introduce randomness in the AQM reduced if network devices introduce randomness in the AQM
functions that trigger congestion avoidance at the sending host. functions that trigger congestion avoidance at the sending host.
2.1. AQM and Multiple Queues 2.1. AQM and Multiple Queues
A network device may use per-flow or per-class queuing with a A network device may use per-flow or per-class queueing with a
scheduling algorithm to either prioritize certain applications or scheduling algorithm to either prioritize certain applications or
classes of traffic, limit the rate of transmission, or to provide classes of traffic, limit the rate of transmission, or provide
isolation between different traffic flows within a common class. For isolation between different traffic flows within a common class. For
example, a router may maintain per-flow state to achieve general example, a router may maintain per-flow state to achieve general
fairness by a per-flow scheduling algorithm such as various forms of fairness by a per-flow scheduling algorithm such as various forms of
Fair Queueing (FQ) [Dem90] [Sut99], including Weighted Fair Queuing Fair Queueing (FQ) [Dem90] [Sut99], including Weighted Fair Queueing
(WFQ), Stochastic Fairness Queueing (SFQ) [McK90] Deficit Round Robin (WFQ), Stochastic Fairness Queueing (SFQ) [McK90], Deficit Round
(DRR) [Shr96], [Nic12], and/or a Class-Based Queue scheduling Robin (DRR) [Shr96] [Nic12], and/or a Class-Based Queue scheduling
algorithm such as CBQ [Floyd95]. Hierarchical queues may also be algorithm such as CBQ [Floyd95]. Hierarchical queues may also be
used e.g., as a part of a Hierarchical Token Bucket (HTB), or used, e.g., as a part of a Hierarchical Token Bucket (HTB) or
Hierarchical Fair Service Curve (HFSC) [Sto97]. These methods are Hierarchical Fair Service Curve (HFSC) [Sto97]. These methods are
also used to realize a range of Quality of Service (QoS) behaviours also used to realize a range of Quality of Service (QoS) behaviors
designed to meet the need of traffic classes (e.g. using the designed to meet the need of traffic classes (e.g., using the
integrated or differentiated service models). integrated or differentiated service models).
AQM is needed even for network devices that use per-flow or per-class AQM is needed even for network devices that use per-flow or per-class
queuing, because scheduling algorithms by themselves do not control queueing, because scheduling algorithms by themselves do not control
the overall queue size or the size of individual queues. AQM the overall queue size or the sizes of individual queues. AQM
mechanisms might need to control the overall queue sizes, to ensure mechanisms might need to control the overall queue sizes to ensure
that arriving bursts can be accommodated without dropping packets. that arriving bursts can be accommodated without dropping packets.
AQM should also be used to control the queue size for each individual AQM should also be used to control the queue size for each individual
flow or class, so that they do not experience unnecessarily high flow or class, so that they do not experience unnecessarily high
delay. Using a combination of AQM and scheduling between multiple delay. Using a combination of AQM and scheduling between multiple
queues has been shown to offer good results in experimental and some queues has been shown to offer good results in experimental use and
types of operational use. some types of operational use.
In short, scheduling algorithms and queue management should be seen In short, scheduling algorithms and queue management should be seen
as complementary, not as replacements for each other. as complementary, not as replacements for each other.
2.2. AQM and Explicit Congestion Marking (ECN) 2.2. AQM and Explicit Congestion Marking (ECN)
An AQM method may use Explicit Congestion Notification (ECN) An AQM method may use Explicit Congestion Notification (ECN)
[RFC3168] instead of dropping to mark packets under mild or moderate [RFC3168] instead of dropping to mark packets under mild or moderate
congestion. ECN-marking can allow a network device to signal congestion. ECN-marking can allow a network device to signal
congestion at a point before a transport experiences congestion loss congestion at a point before a transport experiences congestion loss
or additional queuing delay [ECN-Benefit]. Section 4.2.1 describes or additional queueing delay [ECN-Benefit]. Section 4.2.1 describes
some of the benefits of using ECN with AQM. some of the benefits of using ECN with AQM.
2.3. AQM and Buffer Size 2.3. AQM and Buffer Size
It is important to differentiate the choice of buffer size for a It is important to differentiate the choice of buffer size for a
queue in a switch/router or other network device, and the queue in a switch/router or other network device, and the
threshold(s) and other parameters that determine how and when an AQM threshold(s) and other parameters that determine how and when an AQM
algorithm operates. The optimum buffer size is a function of algorithm operates. The optimum buffer size is a function of
operational requirements and should generally be sized to be operational requirements and should generally be sized to be
sufficient to buffer the largest normal traffic burst that is sufficient to buffer the largest normal traffic burst that is
expected. This size depends on the number and burstiness of traffic expected. This size depends on the amount and burstiness of traffic
arriving at the queue and the rate at which traffic leaves the queue. arriving at the queue and the rate at which traffic leaves the queue.
One objective of AQM is to minimize the effect of lock-out, where one One objective of AQM is to minimize the effect of lock-out, where one
flow prevents other flows from effectively gaining capacity. This flow prevents other flows from effectively gaining capacity. This
need can be illustrated by a simple example of drop-tail queuing when need can be illustrated by a simple example of drop-tail queueing
a new TCP flow injects packets into a queue that happens to be almost when a new TCP flow injects packets into a queue that happens to be
full. A TCP flow's congestion control algorithm [RFC5681] increases almost full. A TCP flow's congestion control algorithm [RFC5681]
the flow rate to maximize its effective window. This builds a queue increases the flow rate to maximize its effective window. This
in the network, inducing latency to the flow and other flows that builds a queue in the network, inducing latency in the flow and other
share this queue. Once a drop-tail queue fills, there will also be flows that share this queue. Once a drop-tail queue fills, there
loss. A new flow, sending its initial burst, has an enhanced will also be loss. A new flow, sending its initial burst, has an
probability of filling the remaining queue and dropping packets. As enhanced probability of filling the remaining queue and dropping
a result, the new flow can be effectively prevented from effectively packets. As a result, the new flow can be prevented from effectively
sharing the queue for a period of many RTTs. In contrast, AQM can sharing the queue for a period of many RTTs. In contrast, AQM can
minimize the mean queue depth and therefore reducing the probability minimize the mean queue depth and therefore reduce the probability
that competing sessions can materially prevent each other from that competing sessions can materially prevent each other from
performing well. performing well.
AQM frees a designer from having to limit the buffer space assigned AQM frees a designer from having to limit the buffer space assigned
to a queue to achieve acceptable performance, allowing allocation of to a queue to achieve acceptable performance, allowing allocation of
sufficient buffering to satisfy the needs of the particular traffic sufficient buffering to satisfy the needs of the particular traffic
pattern. Different types of traffic and deployment scenarios will pattern. Different types of traffic and deployment scenarios will
lead to different requirements. The choice of AQM algorithm and lead to different requirements. The choice of AQM algorithm and
associated parameters is therefore a function of the way in which associated parameters is therefore a function of the way in which
congestion is experienced and the required reaction to achieve congestion is experienced and the required reaction to achieve
acceptable performance. This latter is the primary topic of the acceptable performance. The latter is the primary topic of the
following sections. following sections.
3. Managing Aggressive Flows 3. Managing Aggressive Flows
One of the keys to the success of the Internet has been the One of the keys to the success of the Internet has been the
congestion avoidance mechanisms of TCP. Because TCP "backs off" congestion avoidance mechanisms of TCP. Because TCP "backs off"
during congestion, a large number of TCP connections can share a during congestion, a large number of TCP connections can share a
single, congested link in such a way that link bandwidth is shared single, congested link in such a way that link bandwidth is shared
reasonably equitably among similarly situated flows. The equitable reasonably equitably among similarly situated flows. The equitable
sharing of bandwidth among flows depends on all flows running sharing of bandwidth among flows depends on all flows running
compatible congestion avoidance algorithms, i.e., methods conformant compatible congestion avoidance algorithms, i.e., methods conformant
with the current TCP specification [RFC5681]. with the current TCP specification [RFC5681].
In this document a flow is known as "TCP-friendly" when it has a In this document, a flow is known as "TCP-friendly" when it has a
congestion response that approximates the average response expected congestion response that approximates the average response expected
of a TCP flow. One example method of a TCP-friendly scheme is the of a TCP flow. One example method of a TCP-friendly scheme is the
TCP-Friendly Rate Control algorithm [RFC5348]. In this document, the TCP-Friendly Rate Control algorithm [RFC5348]. In this document, the
term is used more generally to describe this and other algorithms term is used more generally to describe this and other algorithms
that meet these goals. that meet these goals.
There are a variety of types of network flow. Some convenient There are a variety of types of network flow. Some convenient
classes that describe flows are: (1) TCP Friendly flows, (2) classes that describe flows are: (1) TCP-friendly flows, (2)
unresponsive flows, i.e., flows that do not slow down when congestion unresponsive flows, i.e., flows that do not slow down when congestion
occurs, and (3) flows that are responsive but are less responsive to occurs, and (3) flows that are responsive but are less responsive to
congestion than TCP. The last two classes contain more aggressive congestion than TCP. The last two classes contain more aggressive
flows that can pose significant threats to Internet performance. flows that can pose significant threats to Internet performance.
1. TCP-Friendly flows 1. TCP-friendly flows
A TCP-friendly flow responds to congestion notification within a A TCP-friendly flow responds to congestion notification within a
small number of path Round Trip Times (RTT), and in steady-state small number of path RTTs, and in steady-state it uses no more
it uses no more capacity than a conformant TCP running under capacity than a conformant TCP running under comparable
comparable conditions (drop rate, RTT, packet size, etc.). This conditions (drop rate, RTT, packet size, etc.). This is
is described in the remainder of the document. described in the remainder of the document.
2. Non-Responsive Flows 2. Non-responsive flows
A flow that does not adjust its rate in response to congestion A non-responsive flow does not adjust its rate in response to
notification within a small number of path RTTs, can also use congestion notification within a small number of path RTTs; it
more capacity than a conformant TCP running under comparable can also use more capacity than a conformant TCP running under
conditions. There is a growing set of applications whose comparable conditions. There is a growing set of applications
congestion avoidance algorithms are inadequate or nonexistent whose congestion avoidance algorithms are inadequate or
(i.e., a flow that does not throttle its sending rate when it nonexistent (i.e., a flow that does not throttle its sending rate
experiences congestion). when it experiences congestion).
The User Datagram Protocol (UDP) [RFC0768] provides a minimal, The User Datagram Protocol (UDP) [RFC768] provides a minimal,
best-effort transport to applications and upper-layer protocols best-effort transport to applications and upper-layer protocols
(both simply called "applications" in the remainder of this (both simply called "applications" in the remainder of this
document) and does not itself provide mechanisms to prevent document) and does not itself provide mechanisms to prevent
congestion collapse and establish a degree of fairness [RFC5405]. congestion collapse or establish a degree of fairness [RFC5405].
Examples that use UDP include some streaming applications for Examples that use UDP include some streaming applications for
packet voice and video, and some multicast bulk data transport. packet voice and video, and some multicast bulk data transport.
Other traffic, when aggregated may also become unresponsive to Other traffic, when aggregated, may also become unresponsive to
congestion notification. If no action is taken, such congestion notification. If no action is taken, such
unresponsive flows could lead to a new congestion collapse unresponsive flows could lead to a new congestion collapse
[RFC2914]. Some applications can even increase their traffic [RFC2914]. Some applications can even increase their traffic
volume in response to congestion (e.g. by adding forward error volume in response to congestion (e.g., by adding Forward Error
correction when loss is experienced), with the possibility that Correction when loss is experienced), with the possibility that
they contribute to congestion collapse. they contribute to congestion collapse.
In general, applications need to incorporate effective congestion In general, applications need to incorporate effective congestion
avoidance mechanisms [RFC5405]. Research continues to be needed avoidance mechanisms [RFC5405]. Research continues to be needed
to identify and develop ways to accomplish congestion avoidance to identify and develop ways to accomplish congestion avoidance
for presently unresponsive applications. Network devices need to for presently unresponsive applications. Network devices need to
be able to protect themselves against unresponsive flows, and be able to protect themselves against unresponsive flows, and
mechanisms to accomplish this must be developed and deployed. mechanisms to accomplish this must be developed and deployed.
Deployment of such mechanisms would provide an incentive for all Deployment of such mechanisms would provide an incentive for all
applications to become responsive by either using a congestion- applications to become responsive by either using a congestion-
controlled transport (e.g. TCP, SCTP [RFC4960] and DCCP controlled transport (e.g., TCP, SCTP [RFC4960], and DCCP
[RFC4340].) or by incorporating their own congestion control in [RFC4340]) or incorporating their own congestion control in the
the application [RFC5405], [RFC6679]. application [RFC5405] [RFC6679].
3. Transport Flows that are less responsive than TCP 3. Transport flows that are less responsive than TCP
A second threat is posed by transport protocol implementations A second threat is posed by transport protocol implementations
that are responsive to congestion, but, either deliberately or that are responsive to congestion, but, either deliberately or
through faulty implementation, reduce less than a TCP flow would through faulty implementation, reduce the effective window less
have done in response to congestion. This covers a spectrum of than a TCP flow would have done in response to congestion. This
behaviours between (1) and (2). If applications are not covers a spectrum of behaviors between (1) and (2). If
sufficiently responsive to congestion signals, they may gain an applications are not sufficiently responsive to congestion
unfair share of the available network capacity. signals, they may gain an unfair share of the available network
capacity.
For example, the popularity of the Internet has caused a For example, the popularity of the Internet has caused a
proliferation in the number of TCP implementations. Some of proliferation in the number of TCP implementations. Some of
these may fail to implement the TCP congestion avoidance these may fail to implement the TCP congestion avoidance
mechanisms correctly because of poor implementation. Others may mechanisms correctly because of poor implementation. Others may
deliberately be implemented with congestion avoidance algorithms deliberately be implemented with congestion avoidance algorithms
that are more aggressive in their use of capacity than other TCP that are more aggressive in their use of capacity than other TCP
implementations; this would allow a vendor to claim to have a implementations; this would allow a vendor to claim to have a
"faster TCP". The logical consequence of such implementations "faster TCP". The logical consequence of such implementations
would be a spiral of increasingly aggressive TCP implementations, would be a spiral of increasingly aggressive TCP implementations,
skipping to change at page 13, line 42 skipping to change at page 15, line 4
that are more aggressive in their use of capacity than other TCP that are more aggressive in their use of capacity than other TCP
implementations; this would allow a vendor to claim to have a implementations; this would allow a vendor to claim to have a
"faster TCP". The logical consequence of such implementations "faster TCP". The logical consequence of such implementations
would be a spiral of increasingly aggressive TCP implementations, would be a spiral of increasingly aggressive TCP implementations,
leading back to the point where there is effectively no leading back to the point where there is effectively no
congestion avoidance and the Internet is chronically congested. congestion avoidance and the Internet is chronically congested.
Another example could be an RTP/UDP video flow that uses an Another example could be an RTP/UDP video flow that uses an
adaptive codec, but responds incompletely to indications of adaptive codec, but responds incompletely to indications of
congestion or responds over an excessively long time period. congestion or responds over an excessively long time period.
Such flows are unlikely to be responsive to congestion signals in Such flows are unlikely to be responsive to congestion signals in
a timeframe comparable to a small number of end-to-end a time frame comparable to a small number of end-to-end
transmission delays. However, over a longer timescale, perhaps transmission delays. However, over a longer timescale, perhaps
seconds in duration, they could moderate their speed, or increase seconds in duration, they could moderate their speed, or increase
their speed if they determine capacity to be available. their speed if they determine capacity to be available.
Tunneled traffic aggregates carrying multiple (short) TCP flows Tunneled traffic aggregates carrying multiple (short) TCP flows
can be more aggressive than standard bulk TCP. Applications can be more aggressive than standard bulk TCP. Applications
(e.g., web browsers primarily supporting HTTP 1.1 and peer-to- (e.g., web browsers primarily supporting HTTP 1.1 and peer-to-
peer file-sharing) have exploited this by opening multiple peer file-sharing) have exploited this by opening multiple
connections to the same endpoint. connections to the same endpoint.
Lastly, some applications (e.g., web browsers primarily Lastly, some applications (e.g., web browsers primarily
supporting HTTP 1.1) open a large numbers of succesive short TCP supporting HTTP 1.1) open a large numbers of successive short TCP
flows for a single session. This can lead to each individual flows for a single session. This can lead to each individual
flow spending the majority of time in the exponential TCP slow flow spending the majority of time in the exponential TCP slow
start phase, rather than in TCP congestion avoidance. The start phase, rather than in TCP congestion avoidance. The
resulting traffic aggregate can therefore be much less responsive resulting traffic aggregate can therefore be much less responsive
than a single standard TCP flow. than a single standard TCP flow.
The projected increase in the fraction of total Internet traffic for The projected increase in the fraction of total Internet traffic for
more aggressive flows in classes 2 and 3 could pose a threat to the more aggressive flows in classes 2 and 3 could pose a threat to the
performance of the future Internet. There is therefore an urgent performance of the future Internet. There is therefore an urgent
need for measurements of current conditions and for further research need for measurements of current conditions and for further research
into the ways of managing such flows. This raises many difficult into the ways of managing such flows. This raises many difficult
issues in finding methods with an acceptable overhead cost that can issues in finding methods with an acceptable overhead cost that can
identify and isolate unresponsive flows or flows that are less identify and isolate unresponsive flows or flows that are less
responsive than TCP. Finally, there is as yet little measurement or responsive than TCP. Finally, there is as yet little measurement or
simulation evidence available about the rate at which these threats simulation evidence available about the rate at which these threats
are likely to be realized, or about the expected benefit of are likely to be realized or about the expected benefit of algorithms
algorithms for managing such flows. for managing such flows.
Another topic requiring consideration is the appropriate granularity Another topic requiring consideration is the appropriate granularity
of a "flow" when considering a queue management method. There are a of a "flow" when considering a queue management method. There are a
few "natural" answers: 1) a transport (e.g.,TCP or UDP) flow (source few "natural" answers: 1) a transport (e.g., TCP or UDP) flow (source
address/port, destination address/port, protocol); 2) Differentiated address/port, destination address/port, protocol); 2) Differentiated
Services Code Point, DSCP; 3) a source/destination host pair (IP Services Code Point, DSCP; 3) a source/destination host pair (IP
address); 4) a given source host or a given destination host, or address); 4) a given source host or a given destination host, or
various combinations of the above; 5) a subscriber or site receiving various combinations of the above; 5) a subscriber or site receiving
the Internet service (enterprise or residential). the Internet service (enterprise or residential).
The source/destination host pair gives an appropriate granularity in The source/destination host pair gives an appropriate granularity in
many circumstances, However, different vendors/providers use many circumstances. However, different vendors/providers use
different granularities for defining a flow (as a way of different granularities for defining a flow (as a way of
"distinguishing" themselves from one another), and different "distinguishing" themselves from one another), and different
granularities may be chosen for different places in the network. It granularities may be chosen for different places in the network. It
may be the case that the granularity is less important than the fact may be the case that the granularity is less important than the fact
that a network device needs to be able to deal with more unresponsive that a network device needs to be able to deal with more unresponsive
flows at *some* granularity. The granularity of flows for congestion flows at *some* granularity. The granularity of flows for congestion
management is, at least in part, a question of policy that needs to management is, at least in part, a question of policy that needs to
be addressed in the wider IETF community. be addressed in the wider IETF community.
4. Conclusions and Recommendations 4. Conclusions and Recommendations
The IRTF, in publishing [RFC2309], and the IETF in subsequent The IRTF, in producing [RFC2309], and the IETF in subsequent
discussion, has developed a set of specific recommendations regarding discussion, have developed a set of specific recommendations
the implementation and operational use of AQM procedures. The regarding the implementation and operational use of AQM procedures.
recommendations provided by this document are summarised as: The recommendations provided by this document are summarized as:
1. Network devices SHOULD implement some AQM mechanism to manage 1. Network devices SHOULD implement some AQM mechanism to manage
queue lengths, reduce end-to-end latency, and avoid lock-out queue lengths, reduce end-to-end latency, and avoid lock-out
phenomena within the Internet. phenomena within the Internet.
2. Deployed AQM algorithms SHOULD support Explicit Congestion 2. Deployed AQM algorithms SHOULD support Explicit Congestion
Notification (ECN) as well as loss to signal congestion to Notification (ECN) as well as loss to signal congestion to
endpoints. endpoints.
3. AQM algorithms SHOULD NOT require tuning of initial or 3. AQM algorithms SHOULD NOT require tuning of initial or
configuration parameters in common use cases. configuration parameters in common use cases.
4. AQM algorithms SHOULD respond to measured congestion, not 4. AQM algorithms SHOULD respond to measured congestion, not
application profiles. application profiles.
5. AQM algorithms SHOULD NOT interpret specific transport protocol 5. AQM algorithms SHOULD NOT interpret specific transport protocol
behaviours. behaviors.
6. Transport protocol congestion control algorithms SHOULD maximize 6. Congestion control algorithms for transport protocols SHOULD
their use of available capacity (when there is data to send) maximize their use of available capacity (when there is data to
without incurring undue loss or undue round trip delay. send) without incurring undue loss or undue round-trip delay.
7. Research, engineering, and measurement efforts are needed 7. Research, engineering, and measurement efforts are needed
regarding the design of mechanisms to deal with flows that are regarding the design of mechanisms to deal with flows that are
unresponsive to congestion notification or are responsive, but unresponsive to congestion notification or are responsive, but
are more aggressive than present TCP. are more aggressive than present TCP.
These recommendations are expressed using the word "SHOULD". This is These recommendations are expressed using the word "SHOULD". This is
in recognition that there may be use cases that have not been in recognition that there may be use cases that have not been
envisaged in this document in which the recommendation does not envisaged in this document in which the recommendation does not
apply. Therefore, care should be taken in concluding that one's use apply. Therefore, care should be taken in concluding that one's use
case falls in that category; during the life of the Internet, such case falls in that category; during the life of the Internet, such
use cases have been rarely if ever observed and reported. To the use cases have been rarely, if ever, observed and reported. To the
contrary, available research [Choi04] says that even high speed links contrary, available research [Choi04] says that even high-speed links
in network cores that are normally very stable in depth and behavior in network cores that are normally very stable in depth and behavior
experience occasional issues that need moderation. The experience occasional issues that need moderation. The
recommendations are detailed in the following sections. recommendations are detailed in the following sections.
4.1. Operational deployments SHOULD use AQM procedures 4.1. Operational Deployments SHOULD Use AQM Procedures
AQM procedures are designed to minimize the delay and buffer AQM procedures are designed to minimize the delay and buffer
exhaustion induced in the network by queues that have filled as a exhaustion induced in the network by queues that have filled as a
result of host behavior. Marking and loss behaviors provide a signal result of host behavior. Marking and loss behaviors provide a signal
that buffers within network devices are becoming unnecessarily full, that buffers within network devices are becoming unnecessarily full
and that the sender would do well to moderate its behavior. and that the sender would do well to moderate its behavior.
The use of scheduling mechanisms, such as priority queuing, classful The use of scheduling mechanisms, such as priority queueing, classful
queuing, and fair queuing, is often effective in networks to help a queueing, and fair queueing, is often effective in networks to help a
network serve the needs of a range of applications. Network network serve the needs of a range of applications. Network
operators can use these methods to manage traffic passing a choke operators can use these methods to manage traffic passing a choke
point. This is discussed in [RFC2474] and [RFC2475]. When point. This is discussed in [RFC2474] and [RFC2475]. When
scheduling is used AQM should be applied across the classes or flows scheduling is used, AQM should be applied across the classes or flows
as well as within each class or flow: as well as within each class or flow:
o AQM mechanisms need to control the overall queue sizes, to ensure o AQM mechanisms need to control the overall queue sizes to ensure
that arriving bursts can be accommodated without dropping packets. that arriving bursts can be accommodated without dropping packets.
o AQM mechanisms need to allow combination with other mechanisms, o AQM mechanisms need to allow combination with other mechanisms,
such as scheduling, to allow implementation of policies for such as scheduling, to allow implementation of policies for
providing fairness between different flows. providing fairness between different flows.
o AQM should be used to control the queue size for each individual o AQM should be used to control the queue size for each individual
flow or class, so that they do not experience unnecessarily high flow or class, so that they do not experience unnecessarily high
delay. delay.
4.2. Signaling to the transport endpoints 4.2. Signaling to the Transport Endpoints
There are a number of ways a network device may signal to the end There are a number of ways a network device may signal to the
point that the network is becoming congested and trigger a reduction endpoint that the network is becoming congested and trigger a
in rate. The signalling methods include: reduction in rate. The signaling methods include:
o Delaying transport segments (packets) in flight, such as in a o Delaying transport segments (packets) in flight, such as in a
queue. queue.
o Dropping transport segments (packets) in transit. o Dropping transport segments (packets) in transit.
o Marking transport segments (packets), such as using Explicit o Marking transport segments (packets), such as using Explicit
Congestion Control[RFC3168] [RFC4301] [RFC4774] [RFC6040] Congestion Control [RFC3168] [RFC4301] [RFC4774] [RFC6040]
[RFC6679]. [RFC6679].
Increased network latency is used as an implicit signal of Increased network latency is used as an implicit signal of
congestion. E.g., in TCP additional delay can affect ACK Clocking congestion. For example, in TCP, additional delay can affect ACK
and has the result of reducing the rate of transmission of new data. clocking and has the result of reducing the rate of transmission of
In the Real Time Protocol (RTP), network latency impacts the RTCP- new data. In the Real-time Transport Protocol (RTP), network latency
reported RTT and increased latency can trigger a sender to adjust its impacts the RTCP-reported RTT, and increased latency can trigger a
rate. Methods such as Low Extra Delay Background Transport (LEDBAT) sender to adjust its rate. Methods such as Low Extra Delay
[RFC6817] assume increased latency as a primary signal of congestion. Background Transport (LEDBAT) [RFC6817] assume increased latency as a
Appropriate use of delay-based methods and the implications of AQM primary signal of congestion. Appropriate use of delay-based methods
presently remains an area for further research. and the implications of AQM presently remain an area for further
research.
It is essential that all Internet hosts respond to loss [RFC5681], It is essential that all Internet hosts respond to loss [RFC5681]
[RFC5405][RFC4960][RFC4340]. Packet dropping by network devices that [RFC5405] [RFC4960] [RFC4340]. Packet dropping by network devices
are under load has two effects: It protects the network, which is the that are under load has two effects: It protects the network, which
primary reason that network devices drop packets. The detection of is the primary reason that network devices drop packets. The
loss also provides a signal to a reliable transport (e.g., TCP, SCTP) detection of loss also provides a signal to a reliable transport
that there is potential congestion using a pragmatic heuristic; "when (e.g., TCP, SCTP) that there is incipient congestion, using a
the network discards a message in flight, it may imply the presence pragmatic but ambiguous heuristic. Whereas, when the network
of faulty equipment or media in a path, and it may imply the presence discards a message in flight, the loss may imply the presence of
of congestion. To be conservative, a transport must assume it may be faulty equipment or media in a path, or it may imply the presence of
the latter." Applications using unreliable transports (e.g.,using congestion. To be conservative, a transport must assume it may be
UDP) need to similarly react to loss [RFC5405] the latter. Applications using unreliable transports (e.g., using
UDP) need to similarly react to loss [RFC5405].
Network devices SHOULD use an AQM algorithm to measure local Network devices SHOULD use an AQM algorithm to measure local
congestion and to determine the packets to mark or drop so that the congestion and to determine the packets to mark or drop so that the
congestion is managed. congestion is managed.
In general, dropping multiple packets from the same sessions in the In general, dropping multiple packets from the same sessions in the
same RTT is ineffective, and can reduce throughput. Also, dropping same RTT is ineffective and can reduce throughput. Also, dropping or
or marking packets from multiple sessions simultaneously can have the marking packets from multiple sessions simultaneously can have the
effect of synchronizing them, resulting in increasing peaks and effect of synchronizing them, resulting in increasing peaks and
troughs in the subsequent traffic load. Hence, AQM algorithms SHOULD troughs in the subsequent traffic load. Hence, AQM algorithms SHOULD
randomize dropping in time, to reduce the probability that congestion randomize dropping in time, to reduce the probability that congestion
indications are only experienced by a small proportion of the active indications are only experienced by a small proportion of the active
flows. flows.
Loss due to dropping also has an effect on the efficiency of a flow Loss due to dropping also has an effect on the efficiency of a flow
and can significantly impact some classes of application. In and can significantly impact some classes of application. In
reliable transports the dropped data must be subsequently reliable transports, the dropped data must be subsequently
retransmitted. While other applications/transports may adapt to the retransmitted. While other applications/transports may adapt to the
absence of lost data, this still implies inefficient use of available absence of lost data, this still implies inefficient use of available
capacity and the dropped traffic can affect other flows. Hence, capacity, and the dropped traffic can affect other flows. Hence,
congestion signalling by loss is not entirely positive; it is a congestion signaling by loss is not entirely positive; it is a
necessary evil. necessary evil.
4.2.1. AQM and ECN 4.2.1. AQM and ECN
Explicit Congestion Notification (ECN) [RFC4301] [RFC4774] [RFC6040] Explicit Congestion Notification (ECN) [RFC4301] [RFC4774] [RFC6040]
[RFC6679] is a network-layer function that allows a transport to [RFC6679] is a network-layer function that allows a transport to
receive network congestion information from a network device without receive network congestion information from a network device without
incurring the unintended consequences of loss. ECN includes both incurring the unintended consequences of loss. ECN includes both
transport mechanisms and functions implemented in network devices, transport mechanisms and functions implemented in network devices;
the latter rely upon using AQM to decider when and whether to ECN- the latter rely upon using AQM to decide when and whether to ECN-
mark. mark.
Congestion for ECN-capable transports is signalled by a network Congestion for ECN-capable transports is signaled by a network device
device setting the "Congestion Experienced (CE)" codepoint in the IP setting the "Congestion Experienced (CE)" codepoint in the IP header.
header. This codepoint is noted by the remote receiving end point This codepoint is noted by the remote receiving endpoint and signaled
and signalled back to the sender using a transport protocol back to the sender using a transport protocol mechanism, allowing the
mechanism, allowing the sender to trigger timely congestion control. sender to trigger timely congestion control. The decision to set the
The decision to set the CE codepoint requires an AQM algorithm CE codepoint requires an AQM algorithm configured with a threshold.
configured with a threshold. Non-ECN capable flows (the default) are Non-ECN capable flows (the default) are dropped under congestion.
dropped under congestion.
Network devices SHOULD use an AQM algorithm that marks ECN-capable Network devices SHOULD use an AQM algorithm that marks ECN-capable
traffic when making decisions about the response to congestion. traffic when making decisions about the response to congestion.
Network devices need to implement this method by marking ECN-capable Network devices need to implement this method by marking ECN-capable
traffic or by dropping non-ECN-capable traffic. traffic or by dropping non-ECN-capable traffic.
Safe deployment of ECN requires that network devices drop excessive Safe deployment of ECN requires that network devices drop excessive
traffic, even when marked as originating from an ECN-capable traffic, even when marked as originating from an ECN-capable
transport. This is a necessary safety precaution because: transport. This is a necessary safety precaution because:
1. A non-conformant, broken or malicious receiver could conceal an 1. A non-conformant, broken, or malicious receiver could conceal an
ECN mark, and not report this to the sender; ECN mark and not report this to the sender;
2. A non-conformant, broken or malicious sender could ignore a 2. A non-conformant, broken, or malicious sender could ignore a
reported ECN mark, as it could ignore a loss without using ECN; reported ECN mark, as it could ignore a loss without using ECN;
3. A malfunctioning or non-conforming network device may "hide" an 3. A malfunctioning or non-conforming network device may "hide" an
ECN mark (or fail to correctly set the ECN codepoint at an egress ECN mark (or fail to correctly set the ECN codepoint at an egress
of a network tunnel). of a network tunnel).
In normal operation, such cases should be very uncommon, however In normal operation, such cases should be very uncommon; however,
overload protection is desirable to protect traffic from overload protection is desirable to protect traffic from
misconfigured or malicious use of ECN (e.g., a denial-of-service misconfigured or malicious use of ECN (e.g., a denial-of-service
attack that generates ECN-capable traffic that is unresponsive to CE- attack that generates ECN-capable traffic that is unresponsive to CE-
marking). marking).
An AQM algorithm that supports ECN needs to define the threshold and When ECN is added to a scheme, the ECN support MAY define a separate
algorithm for ECN-marking. This threshold MAY differ from that used set of parameters from those used for controlling packet drop. The
for dropping packets that are not marked as ECN-capable, and SHOULD AQM algorithm SHOULD still auto-tune these ECN-specific parameters.
be configurable. These parameters SHOULD also be manually configurable.
Network devices SHOULD use an algorithm to drop excessive traffic Network devices SHOULD use an algorithm to drop excessive traffic
(e.g., at some level above the threshold for CE-marking), even when (e.g., at some level above the threshold for CE-marking), even when
the packets are marked as originating from an ECN-capable transport. the packets are marked as originating from an ECN-capable transport.
4.3. AQM algorithm deployment SHOULD NOT require operational tuning 4.3. AQM Algorithm Deployment SHOULD NOT Require Operational Tuning
A number of AQM algorithms have been proposed. Many require some A number of AQM algorithms have been proposed. Many require some
form of tuning or setting of parameters for initial network form of tuning or setting of parameters for initial network
conditions. This can make these algorithms difficult to use in conditions. This can make these algorithms difficult to use in
operational networks. operational networks.
AQM algorithms need to consider both "initial conditions" and AQM algorithms need to consider both "initial conditions" and
"operational conditions". The former includes values that exist "operational conditions". The former includes values that exist
before any experience is gathered about the use of the algorithm, before any experience is gathered about the use of the algorithm,
such as the configured speed of interface, support for full duplex such as the configured speed of interface, support for full-duplex
communication, interface MTU and other properties of the link. The communication, interface MTU, and other properties of the link.
latter includes information observed from monitoring the size of the Other properties include information observed from monitoring the
queue, experienced queueing delay, rate of packet discard, etc. size of the queue, the queueing delay experienced, rate of packet
discard, etc.
This document therefore specifies that AQM algorithms that are This document therefore specifies that AQM algorithms that are
proposed for deployment in the Internet have the following proposed for deployment in the Internet have the following
properties: properties:
o AQM algorithm deployment SHOULD NOT require tuning. An algorithm o AQM algorithm deployment SHOULD NOT require tuning. An algorithm
MUST provide a default behaviour that auto-tunes to a reasonable MUST provide a default behavior that auto-tunes to a reasonable
performance for typical network operational conditions. This is performance for typical network operational conditions. This is
expected to ease deployment and operation. Initial conditions, expected to ease deployment and operation. Initial conditions,
such as the interface rate and MTU size or other values derived such as the interface rate and MTU size or other values derived
from these, MAY be required by an AQM algorithm. from these, MAY be required by an AQM algorithm.
o MAY support further manual tuning that could improve performance o AQM algorithm deployment MAY support further manual tuning that
in a specific deployed network. Algorithms that lack such could improve performance in a specific deployed network.
variables are acceptable, but if such variables exist, they SHOULD Algorithms that lack such variables are acceptable, but, if such
be externalized (made visible to the operator). Guidance needs to variables exist, they SHOULD be externalized (made visible to the
be provided on the cases where auto-tuning is unlikely to achieve operator). The specification should identify any cases in which
acceptable performance and to identify the set of parameters that auto-tuning is unlikely to achieve acceptable performance and give
can be tuned. For example, the expected response of an algorithm guidance on the parametric adjustments necessary. For example,
may need to be configured to accommodate the largest expected Path the expected response of an algorithm may need to be configured to
RTT, since this value can not be known at initialization. This accommodate the largest expected Path RTT, since this value cannot
guidance is expected to enable the algorithm to be deployed in be known at initialization. This guidance is expected to enable
networks that have specific characteristics (paths with variable/ the algorithm to be deployed in networks that have specific
larger delay; networks where capacity is impacted by interactions characteristics (paths with variable or larger delay, networks
with lower layer mechanisms, etc). where capacity is impacted by interactions with lower-layer
mechanisms, etc).
o MAY provide logging and alarm signals to assist in identifying if o AQM algorithm deployment MAY provide logging and alarm signals to
an algorithm using manual or auto-tuning is functioning as assist in identifying if an algorithm using manual or auto-tuning
expected. (e.g., this could be based on an internal consistency is functioning as expected. (For example, this could be based on
check between input, output, and mark/drop rates over time). This an internal consistency check between input, output, and mark/drop
is expected to encourage deployment by default and allow operators rates over time.) This is expected to encourage deployment by
to identify potential interactions with other network functions. default and allow operators to identify potential interactions
with other network functions.
Hence, self-tuning algorithms are to be preferred. Algorithms Hence, self-tuning algorithms are to be preferred. Algorithms
recommended for general Internet deployment by the IETF need to be recommended for general Internet deployment by the IETF need to be
designed so that they do not require operational (especially manual) designed so that they do not require operational (especially manual)
configuration or tuning. configuration or tuning.
4.4. AQM algorithms SHOULD respond to measured congestion, not 4.4. AQM Algorithms SHOULD Respond to Measured Congestion, Not
application profiles. Application Profiles
Not all applications transmit packets of the same size. Although Not all applications transmit packets of the same size. Although
applications may be characterized by particular profiles of packet applications may be characterized by particular profiles of packet
size this should not be used as the basis for AQM (see next section). size, this should not be used as the basis for AQM (see Section 4.5).
Other methods exist, e.g., Differentiated Services queueing, Pre- Other methods exist, e.g., Differentiated Services queueing, Pre-
Congestion Notification (PCN) [RFC5559], that can be used to Congestion Notification (PCN) [RFC5559], that can be used to
differentiate and police classes of application. Network devices may differentiate and police classes of application. Network devices may
combine AQM with these traffic classification mechanisms and perform combine AQM with these traffic classification mechanisms and perform
AQM only on specific queues within a network device. AQM only on specific queues within a network device.
An AQM algorithm should not deliberately try to prejudice the size of An AQM algorithm should not deliberately try to prejudice the size of
packet that performs best (i.e., Preferentially drop/mark based only packet that performs best (i.e., preferentially drop/mark based only
on packet size). Procedures for selecting packets to mark/drop on packet size). Procedures for selecting packets to drop/mark
SHOULD observe the actual or projected time that a packet is in a SHOULD observe the actual or projected time that a packet is in a
queue (bytes at a rate being an analog to time). When an AQM queue (bytes at a rate being an analog to time). When an AQM
algorithm decides whether to drop (or mark) a packet, it is algorithm decides whether to drop (or mark) a packet, it is
RECOMMENDED that the size of the particular packet should not be RECOMMENDED that the size of the particular packet not be taken into
taken into account [RFC7141]. account [RFC7141].
Applications (or transports) generally know the packet size that they Applications (or transports) generally know the packet size that they
are using and can hence make their judgments about whether to use are using and can hence make their judgments about whether to use
small or large packets based on the data they wish to send and the small or large packets based on the data they wish to send and the
expected impact on the delay or throughput, or other performance expected impact on the delay, throughput, or other performance
parameter. When a transport or application responds to a dropped or parameter. When a transport or application responds to a dropped or
marked packet, the size of the rate reduction should be proportionate marked packet, the size of the rate reduction should be proportionate
to the size of the packet that was sent [RFC7141]. to the size of the packet that was sent [RFC7141].
AQM-enabled system MAY instantiate different instances of an AQM An AQM-enabled system MAY instantiate different instances of an AQM
algorithm to be applied within the same traffic class. Traffic algorithm to be applied within the same traffic class. Traffic
classes may be differentiated based on an Access Control List (ACL), classes may be differentiated based on an Access Control List (ACL),
the packet Differentiated Services Code Point (DSCP) [RFC5559], the packet DSCP [RFC5559], enabling use of the ECN field (i.e., any
enabling use of the ECN field (i.e., any of ECT(0), ECT(1) or of ECT(0), ECT(1) or CE) [RFC3168] [RFC4774], a multi-field (MF)
CE)[RFC3168] [RFC4774], a multi-field (MF) classifier that combines classifier that combines the values of a set of protocol fields
the values of a set of protocol fields (e.g., IP address, transport, (e.g., IP address, transport, ports), or an equivalent codepoint at a
ports) or an equivalent codepoint at a lower layer. This lower layer. This recommendation goes beyond what is defined in RFC
recommendation goes beyond what is defined in RFC 3168, by allowing 3168 by allowing that an implementation MAY use more than one
that an implementation MAY use more than one instance of an AQM instance of an AQM algorithm to handle both ECN-capable and non-ECN-
algorithm to handle both ECN-capable and non-ECN-capable packets. capable packets.
4.5. AQM algorithms SHOULD NOT be dependent on specific transport 4.5. AQM Algorithms SHOULD NOT Be Dependent on Specific Transport
protocol behaviours Protocol Behaviors
In deploying AQM, network devices need to support a range of Internet In deploying AQM, network devices need to support a range of Internet
traffic and SHOULD NOT make implicit assumptions about the traffic and SHOULD NOT make implicit assumptions about the
characteristics desired by the set transports/applications the characteristics desired by the set of transports/applications the
network supports. That is, AQM methods should be opaque to the network supports. That is, AQM methods should be opaque to the
choice of transport and application. choice of transport and application.
AQM algorithms are often evaluated by considering TCP [RFC0793] with AQM algorithms are often evaluated by considering TCP [RFC793] with a
a limited number of applications. Although TCP is the predominant limited number of applications. Although TCP is the predominant
transport in the Internet today, this no longer represents a transport in the Internet today, this no longer represents a
sufficient selection of traffic for verification. There is sufficient selection of traffic for verification. There is
significant use of UDP [RFC0768] in voice and video services, and significant use of UDP [RFC768] in voice and video services, and some
some applications find utility in SCTP [RFC4960] and DCCP [RFC4340]. applications find utility in SCTP [RFC4960] and DCCP [RFC4340].
Hence, AQM algorithms should also demonstrate operation with Hence, AQM algorithms should demonstrate operation with transports
transports other than TCP and need to consider a variety of other than TCP and need to consider a variety of applications. When
applications. Selection of AQM algorithms also needs to consider use selecting AQM algorithms, the use of tunnel encapsulations that may
of tunnel encapsulations that may carry traffic aggregates. carry traffic aggregates needs to be considered.
AQM algorithms SHOULD NOT target or derive implicit assumptions about AQM algorithms SHOULD NOT target or derive implicit assumptions about
the characteristics desired by specific transports/applications. the characteristics desired by specific transports/applications.
Transports and applications need to respond to the congestion signals Transports and applications need to respond to the congestion signals
provided by AQM (i.e., dropping or ECN-marking) in a timely manner provided by AQM (i.e., dropping or ECN-marking) in a timely manner
(within a few RTT at the latest). (within a few RTTs at the latest).
4.6. Interactions with congestion control algorithms 4.6. Interactions with Congestion Control Algorithms
Applications and transports need to react to received implicit or Applications and transports need to react to received implicit or
explicit signals that indicate the presence of congestion. This explicit signals that indicate the presence of congestion. This
section identifies issues that can impact the design of transport section identifies issues that can impact the design of transport
protocols when using paths that use AQM. protocols when using paths that use AQM.
Transport protocols and applications need timely signals of Transport protocols and applications need timely signals of
congestion. The time taken to detect and respond to congestion is congestion. The time taken to detect and respond to congestion is
increased when network devices queue packets in buffers. It can be increased when network devices queue packets in buffers. It can be
difficult to detect tail losses at a higher layer and this may difficult to detect tail losses at a higher layer, and this may
sometimes require transport timers or probe packets to detect and sometimes require transport timers or probe packets to detect and
respond to such loss. Loss patterns may also impact timely respond to such loss. Loss patterns may also impact timely
detection, e.g., the time may be reduced when network devices do not detection, e.g., the time may be reduced when network devices do not
drop long runs of packets from the same flow. drop long runs of packets from the same flow.
A common objective of an elastic transport congestion control A common objective of an elastic transport congestion control
protocol is to allow an application to deliver the maximum rate of protocol is to allow an application to deliver the maximum rate of
data without inducing excessive delays when packets are queued in a data without inducing excessive delays when packets are queued in
buffers within the network. To achieve this, a transport should try buffers within the network. To achieve this, a transport should try
to operate at rate below the inflexion point of the load/delay curve to operate at rate below the inflection point of the load/delay curve
(the bend of what is sometimes called a "hockey-stick" curve) (the bend of what is sometimes called a "hockey stick" curve)
[Jain94]. When the congestion window allows the load to approach [Jain94]. When the congestion window allows the load to approach
this bend, the end-to-end delay starts to rise - a result of this bend, the end-to-end delay starts to rise -- a result of
congestion, as packets probabilistically arrive at non-overlapping congestion, as packets probabilistically arrive at non-overlapping
times. On the one hand, a transport that operates above this point times. On the one hand, a transport that operates above this point
can experience congestion loss and could also trigger operator can experience congestion loss and could also trigger operator
activities, such as those discussed in [RFC6057]. On the other hand, activities, such as those discussed in [RFC6057]. On the other hand,
a flow may achieve both near-maximum throughput and low latency when a flow may achieve both near-maximum throughput and low latency when
it operates close to this knee point, with minimal contribution to it operates close to this knee point, with minimal contribution to
router congestion. Choice of an appropriate rate/congestion window router congestion. Choice of an appropriate rate/congestion window
can therefore significantly impact the loss and delay experienced by can therefore significantly impact the loss and delay experienced by
a flow and will impact other flows that share a common network queue. a flow and will impact other flows that share a common network queue.
Some applications may send less than permitted by the congestion Some applications may send data at a lower rate or keep less segments
control window (or rate). Examples include multimedia codecs that outstanding at any given time. Examples include multimedia codecs
stream at some natural rate (or set of rates) or an application that that stream at some natural rate (or set of rates) or an application
is naturally interactive (e.g., some web applications, interactive that is naturally interactive (e.g., some web applications,
server-based gaming, transaction-based protocols). Such applications interactive server-based gaming, transaction-based protocols). Such
may have different objectives. They may not wish to maximize applications may have different objectives. They may not wish to
throughput, but may desire a lower loss rate or bounded delay. maximize throughput, but may desire a lower loss rate or bounded
delay.
The correct operation of an AQM-enabled network device MUST NOT rely The correct operation of an AQM-enabled network device MUST NOT rely
upon specific transport responses to congestion signals. upon specific transport responses to congestion signals.
4.7. The need for further research 4.7. The Need for Further Research
The second recommendation of [RFC2309] called for further research The second recommendation of [RFC2309] called for further research
into the interaction between network queues and host applications, into the interaction between network queues and host applications,
and the means of signaling between them. This research has occurred, and the means of signaling between them. This research has occurred,
and we as a community have learned a lot. However, we are not done. and we as a community have learned a lot. However, we are not done.
We have learned that the problems of congestion, latency and buffer- We have learned that the problems of congestion, latency, and buffer-
sizing have not gone away, and are becoming more important to many sizing have not gone away and are becoming more important to many
users. A number of self-tuning AQM algorithms have been found that users. A number of self-tuning AQM algorithms have been found that
offer significant advantages for deployed networks. There is also offer significant advantages for deployed networks. There is also
renewed interest in deploying AQM and the potential of ECN. renewed interest in deploying AQM and the potential of ECN.
Traffic patterns can depend on the network deployment scenario, and Traffic patterns can depend on the network deployment scenario, and
Internet research therefore needs to consider the implications of a Internet research therefore needs to consider the implications of a
diverse range of application interactions. This includes ensuring diverse range of application interactions. This includes ensuring
that combinations of mechanisms, as well as combinations of traffic that combinations of mechanisms, as well as combinations of traffic
patterns, do not interact and result in either significantly reduced patterns, do not interact and result in either significantly reduced
flow throughput or significantly increased latency. flow throughput or significantly increased latency.
At the time of writing (in 2015), an obvious example of further At the time of writing (in 2015), an obvious example of further
research is the need to consider the many-to-one communication research is the need to consider the many-to-one communication
patterns found in data centers, known as incast [Ren12], (e.g., patterns found in data centers, known as incast [Ren12], (e.g.,
produced by Map/Reduce applications). Such anlaysis needs to study produced by Map/Reduce applications). Such analysis needs to study
not only each application traffic type, but should also include not only each application traffic type but also combinations of types
combinations of types of traffic. of traffic.
Research also needs to consider the need to extend our taxonomy of Research also needs to consider the need to extend our taxonomy of
transport sessions to include not only "mice" and "elephants", but transport sessions to include not only "mice" and "elephants", but
"lemmings"? Where "Lemmings" are flash crowds of "mice" that the "lemmings". Here, "lemmings" are flash crowds of "mice" that the
network inadvertently tries to signal to as if they were elephant network inadvertently tries to signal to as if they were "elephant"
flows, resulting in head of line blocking in a data center deployment flows, resulting in head-of-line blocking in a data center deployment
scenario. scenario.
Examples of other required research include: Examples of other required research include:
o Research into new AQM and scheduling algorithms. o new AQM and scheduling algorithms
o Appropriate use of delay-based methods and the implications of
AQM.
o Research into suitable algorithms for marking ECN-capable packets
that do not require operational configuration or tuning for common
use.
o Experience in the deployment of ECN alongside AQM. o appropriate use of delay-based methods and the implications of AQM
o Tools for enabling AQM (and ECN) deployment and measuring the o suitable algorithms for marking ECN-capable packets that do not
performance. require operational configuration or tuning for common use
o Methods for mitigating the impact of non-conformant and malicious o experience in the deployment of ECN alongside AQM
flows.
o Research to understand the implications of using new network and o tools for enabling AQM (and ECN) deployment and measuring the
transport methods on applications. performance
Hence, this document therefore reiterates the call of RFC 2309: we o methods for mitigating the impact of non-conformant and malicious
need continuing research as applications develop. flows
5. IANA Considerations o implications on applications of using new network and transport
methods
This memo asks the IANA for no new parameters. Hence, this document reiterates the call of RFC 2309: we need
continuing research as applications develop.
6. Security Considerations 5. Security Considerations
While security is a very important issue, it is largely orthogonal to While security is a very important issue, it is largely orthogonal to
the performance issues discussed in this memo. the performance issues discussed in this memo.
This recommendation requires algorithms to be independent of specific This recommendation requires algorithms to be independent of specific
transport or application behaviors. Therefore a network device does transport or application behaviors. Therefore, a network device does
not require visibility or access to upper layer protocol information not require visibility or access to upper-layer protocol information
to implement an AQM algorithm. This ability to operate in an to implement an AQM algorithm. This ability to operate in an
application-agnostic fashion is therefore an example of a privacy- application-agnostic fashion is an example of a privacy-enhancing
enhancing feature. feature.
Many deployed network devices use queueing methods that allow Many deployed network devices use queueing methods that allow
unresponsive traffic to capture network capacity, denying access to unresponsive traffic to capture network capacity, denying access to
other traffic flows. This could potentially be used as a denial-of- other traffic flows. This could potentially be used as a denial-of-
service attack. This threat could be reduced in network devices that service attack. This threat could be reduced in network devices that
deploy AQM or some form of scheduling. We note, however, that a deploy AQM or some form of scheduling. We note, however, that a
denial-of-service attack that results in unresponsive traffic flows denial-of-service attack that results in unresponsive traffic flows
may be indistinguishable from other traffic flows (e.g., tunnels may be indistinguishable from other traffic flows (e.g., tunnels
carrying aggregates of short flows, high-rate isochronous carrying aggregates of short flows, high-rate isochronous
applications). New methods therefore may remain vulnerable, and this applications). New methods therefore may remain vulnerable, and this
document recommends that ongoing research should consider ways to document recommends that ongoing research consider ways to mitigate
mitigate such attacks. such attacks.
7. Privacy Considerations 6. Privacy Considerations
This document, by itself, presents no new privacy issues. This document, by itself, presents no new privacy issues.
8. Acknowledgements 7. References
The original version of this document describing best current
practice was based on the informational text of [RFC2309]. This was
written by the End-to-End Research Group, which is to say Bob Braden,
Dave Clark, Jon Crowcroft, Bruce Davie, Steve Deering, Deborah
Estrin, Sally Floyd, Van Jacobson, Greg Minshall, Craig Partridge,
Larry Peterson, KK Ramakrishnan, Scott Shenker, John Wroclawski, and
Lixia Zhang. Although there are important differences, many of the
key arguments in the present document remain unchanged from those in
RFC 2309.
The need for an updated document was agreed to in the tsvarea meeting
at IETF 86. This document was reviewed on the aqm@ietf.org list.
Comments were received from Colin Perkins, Richard Scheffenegger,
Dave Taht, John Leslie, David Collier-Brown and many others.
Gorry Fairhurst was in part supported by the European Community under
its Seventh Framework Programme through the Reducing Internet
Transport Latency (RITE) project (ICT-317700).
9. References
9.1. Normative References 7.1. Normative References
[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, March 1997. Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<http://www.rfc-editor.org/info/rfc2119>.
[RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition
of Explicit Congestion Notification (ECN) to IP", RFC of Explicit Congestion Notification (ECN) to IP",
3168, September 2001. RFC 3168, DOI 10.17487/RFC3168, September 2001,
<http://www.rfc-editor.org/info/rfc3168>.
[RFC4301] Kent, S. and K. Seo, "Security Architecture for the [RFC4301] Kent, S. and K. Seo, "Security Architecture for the
Internet Protocol", RFC 4301, December 2005. Internet Protocol", RFC 4301, DOI 10.17487/RFC4301,
December 2005, <http://www.rfc-editor.org/info/rfc4301>.
[RFC4774] Floyd, S., "Specifying Alternate Semantics for the [RFC4774] Floyd, S., "Specifying Alternate Semantics for the
Explicit Congestion Notification (ECN) Field", BCP 124, Explicit Congestion Notification (ECN) Field", BCP 124,
RFC 4774, November 2006. RFC 4774, DOI 10.17487/RFC4774, November 2006,
<http://www.rfc-editor.org/info/rfc4774>.
[RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines [RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines
for Application Designers", BCP 145, RFC 5405, November for Application Designers", BCP 145, RFC 5405, DOI
2008. 10.17487/RFC5405, November 2008,
<http://www.rfc-editor.org/info/rfc5405>.
[RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion [RFC5681] Allman, M., Paxson, V., and E. Blanton, "TCP Congestion
Control", RFC 5681, September 2009. Control", RFC 5681, DOI 10.17487/RFC5681, September 2009,
<http://www.rfc-editor.org/info/rfc5681>.
[RFC6040] Briscoe, B., "Tunnelling of Explicit Congestion [RFC6040] Briscoe, B., "Tunnelling of Explicit Congestion
Notification", RFC 6040, November 2010. Notification", RFC 6040, DOI 10.17487/RFC6040, November
2010, <http://www.rfc-editor.org/info/rfc6040>.
[RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P., [RFC6679] Westerlund, M., Johansson, I., Perkins, C., O'Hanlon, P.,
and K. Carlberg, "Explicit Congestion Notification (ECN) and K. Carlberg, "Explicit Congestion Notification (ECN)
for RTP over UDP", RFC 6679, August 2012. for RTP over UDP", RFC 6679, DOI 10.17487/RFC6679, August
2012, <http://www.rfc-editor.org/info/rfc6679>.
[RFC7141] Briscoe, B. and J. Manner, "Byte and Packet Congestion [RFC7141] Briscoe, B. and J. Manner, "Byte and Packet Congestion
Notification", BCP 41, RFC 7141, February 2014. Notification", BCP 41, RFC 7141, DOI 10.17487/RFC7141,
February 2014, <http://www.rfc-editor.org/info/rfc7141>.
9.2. Informative References 7.2. Informative References
[AQM-WG] "IETF AQM WG", . [AQM-WG] IETF, "Active Queue Management and Packet Scheduling (aqm)
WG", <http://datatracker.ietf.org/wg/aqm/charter/>.
[Bri15] Briscoe, Bob., Brunstrom, Anna., Petlund, Andreas., Hayes, [Bri15] Briscoe, B., Brunstrom, A., Petlund, A., Hayes, D., Ros,
David., Ros, David., Tsang, Ing-Jyh., Gjessing, Stein., D., Tsang, I., Gjessing, S., Fairhurst, G., Griwodz, C.,
Fairhurst, Gorry., Griwodz, Carsten., and Michael. Welzl, and M. Welzl, "Reducing Internet Latency: A Survey of
"Reducing Internet Latency: A Survey of Techniques and Techniques and their Merit", IEEE Communications Surveys &
their Merit, IEEE Communications Surveys & Tutorials", Tutorials, 2015.
2015.
[CONEX] Mathis, M. and B. Briscoe, "The Benefits to Applications [Choi04] Choi, B., Moon, S., Zhang, Z., Papagiannaki, K., and C.
of using Explicit Congestion Notification (ECN)", IETF Diot, "Analysis of Point-To-Point Packet Delay In an
(Work-in-Progress) draft-ietf-conex-abstract-mech, March Operational Network", March 2004.
2014.
[Choi04] Choi, Baek-Young., Moon, Sue., Zhang, Zhi-Li., [CONEX] Mathis, M. and B. Briscoe, "Congestion Exposure (ConEx)
Papagiannaki, K., and C. Diot, "Analysis of Point-To-Point Concepts, Abstract Mechanism and Requirements", Work in
Packet Delay In an Operational Network", March 2004. Progress, draft-ietf-conex-abstract-mech-13, October 2014.
[Dem90] Demers, A., Keshav, S., and S. Shenker, "Analysis and [Dem90] Demers, A., Keshav, S., and S. Shenker, "Analysis and
Simulation of a Fair Queueing Algorithm, Internetworking: Simulation of a Fair Queueing Algorithm, Internetworking:
Research and Experience", SIGCOMM Symposium proceedings on Research and Experience", SIGCOMM Symposium proceedings on
Communications architectures and protocols , 1990. Communications architectures and protocols, 1990.
[ECN-Benefit] [ECN-Benefit]
Welzl, M. and G. Fairhurst, "The Benefits to Applications Fairhurst, G. and M. Welzl, "The Benefits of using
of using Explicit Congestion Notification (ECN)", IETF Explicit Congestion Notification (ECN)", Work in Progress,
(Work-in-Progress) , February 2014. draft-ietf-aqm-ecn-benefits-05, June 2015.
[Flo92] Floyd, S. and V. Jacobsen, "On Traffic Phase Effects in [Flo92] Floyd, S. and V. Jacobsen, "On Traffic Phase Effects in
Packet-Switched Gateways", 1992. Packet-Switched Gateways", 1992,
<http://www.icir.org/floyd/papers/phase.pdf>.
[Flo94] Floyd, S. and V. Jacobsen, "The Synchronization of [Flo94] Floyd, S. and V. Jacobsen, "The Synchronization of
Periodic Routing Messages, Periodic Routing Messages", 1994,
http://ee.lbl.gov/papers/sync_94.pdf", 1994. <http://ee.lbl.gov/papers/sync_94.pdf>.
[Floyd91] Floyd, S., "Connections with Multiple Congested Gateways [Floyd91] Floyd, S., "Connections with Multiple Congested Gateways
in Packet-Switched Networks Part 1: One-way Traffic.", in Packet-Switched Networks Part 1: One-way Traffic.",
Computer Communications Review , October 1991. Computer Communications Review , October 1991.
[Floyd95] Floyd, S. and V. Jacobson, "Link-sharing and Resource [Floyd95] Floyd, S. and V. Jacobson, "Link-sharing and Resource
Management Models for Packet Networks", IEEE/ACM Management Models for Packet Networks", IEEE/ACM
Transactions on Networking , August 1995. Transactions on Networking, August 1995.
[Jacobson88] [Jacobson88]
Jacobson, V., "Congestion Avoidance and Control", SIGCOMM Jacobson, V., "Congestion Avoidance and Control", SIGCOMM
Symposium proceedings on Communications architectures and Symposium proceedings on Communications architectures and
protocols , August 1988. protocols, August 1988.
[Jain94] Jain, Raj., Ramakrishnan, KK., and Chiu. Dah-Ming, [Jain94] Jain, R., Ramakrishnan, KK., and C. Dah-Ming, "Congestion
"Congestion avoidance scheme for computer networks", US avoidance scheme for computer networks", US Patent Office
Patent Office 5377327, December 1994. 5377327, December 1994.
[Lakshman96] [Lakshman96]
Lakshman, TV., Neidhardt, A., and T. Ott, "The Drop From Lakshman, TV., Neidhardt, A., and T. Ott, "The Drop From
Front Strategy in TCP Over ATM and Its Interworking with Front Strategy in TCP Over ATM and Its Interworking with
Other Control Features", IEEE Infocomm , 1996. Other Control Features", IEEE Infocomm, 1996.
[Leland94] [Leland94] Leland, W., Taqqu, M., Willinger, W., and D. Wilson, "On
Leland, W., Taqqu, M., Willinger, W., and D. Wilson, "On
the Self-Similar Nature of Ethernet Traffic (Extended the Self-Similar Nature of Ethernet Traffic (Extended
Version)", IEEE/ACM Transactions on Networking , February Version)", IEEE/ACM Transactions on Networking, February
1994. 1994.
[McK90] McKenney, PE. and G. Varghese, "Stochastic Fairness [McK90] McKenney, PE. and G. Varghese, "Stochastic Fairness
Queuing", Queuing", 1990,
http://www2.rdrop.com/~paulmck/scalability/paper/ <http://www2.rdrop.com/~paulmck/scalability/paper/
sfq.2002.06.04.pdf , 1990. sfq.2002.06.04.pdf>.
[Nic12] Nichols, K., "Controlling Queue Delay", Communications of [Nic12] Nichols, K. and V. Jacobson, "Controlling Queue Delay",
the ACM Vol. 55 No. 11, July, 2012, pp.42-50. , July 2002. Communications of the ACM, Vol. 55, Issue 7, pp. 42-50,
July 2012.
[RFC0768] Postel, J., "User Datagram Protocol", STD 6, RFC 768, [Ren12] Ren, Y., Zhao, Y., and P. Liu, "A survey on TCP Incast in
August 1980. data center networks", International Journal of
Communication Systems, Volumes 27, Issue 8, pages 116-117,
1990.
[RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791, September [RFC768] Postel, J., "User Datagram Protocol", STD 6, RFC 768,
1981. DOI 10.17487/RFC0768, August 1980,
<http://www.rfc-editor.org/info/rfc768>.
[RFC0793] Postel, J., "Transmission Control Protocol", STD 7, RFC [RFC791] Postel, J., "Internet Protocol", STD 5, RFC 791,
793, September 1981. DOI 10.17487/RFC0791, September 1981,
<http://www.rfc-editor.org/info/rfc791>.
[RFC0896] Nagle, J., "Congestion control in IP/TCP internetworks", [RFC793] Postel, J., "Transmission Control Protocol", STD 7,
RFC 896, January 1984. RFC 793, DOI 10.17487/RFC0793, September 1981,
<http://www.rfc-editor.org/info/rfc793>.
[RFC0970] Nagle, J., "On packet switches with infinite storage", RFC [RFC896] Nagle, J., "Congestion Control in IP/TCP Internetworks",
970, December 1985. RFC 896, DOI 10.17487/RFC0896, January 1984,
<http://www.rfc-editor.org/info/rfc896>.
[RFC1122] Braden, R., "Requirements for Internet Hosts - [RFC970] Nagle, J., "On Packet Switches With Infinite Storage",
Communication Layers", STD 3, RFC 1122, October 1989. RFC 970, DOI 10.17487/RFC0970, December 1985,
<http://www.rfc-editor.org/info/rfc970>.
[RFC1633] Braden, B., Clark, D., and S. Shenker, "Integrated [RFC1122] Braden, R., Ed., "Requirements for Internet Hosts -
Services in the Internet Architecture: an Overview", RFC Communication Layers", STD 3, RFC 1122,
1633, June 1994. DOI 10.17487/RFC1122, October 1989,
<http://www.rfc-editor.org/info/rfc1122>.
[RFC1633] Braden, R., Clark, D., and S. Shenker, "Integrated
Services in the Internet Architecture: an Overview",
RFC 1633, DOI 10.17487/RFC1633, June 1994,
<http://www.rfc-editor.org/info/rfc1633>.
[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, April 1998. Internet", RFC 2309, DOI 10.17487/RFC2309, April 1998,
<http://www.rfc-editor.org/info/rfc2309>.
[RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 [RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6
(IPv6) Specification", RFC 2460, December 1998. (IPv6) Specification", RFC 2460, DOI 10.17487/RFC2460,
December 1998, <http://www.rfc-editor.org/info/rfc2460>.
[RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black, [RFC2474] Nichols, K., Blake, S., Baker, F., and D. Black,
"Definition of the Differentiated Services Field (DS "Definition of the Differentiated Services Field (DS
Field) in the IPv4 and IPv6 Headers", RFC 2474, December Field) in the IPv4 and IPv6 Headers", RFC 2474,
1998. DOI 10.17487/RFC2474, December 1998,
<http://www.rfc-editor.org/info/rfc2474>.
[RFC2475] Blake, S., Black, D., Carlson, M., Davies, E., Wang, Z., [RFC2475] Blake, S., Black, D., Carlson, M., Davies, E., Wang, Z.,
and W. Weiss, "An Architecture for Differentiated and W. Weiss, "An Architecture for Differentiated
Services", RFC 2475, December 1998. Services", RFC 2475, DOI 10.17487/RFC2475, December 1998,
<http://www.rfc-editor.org/info/rfc2475>.
[RFC2914] Floyd, S., "Congestion Control Principles", BCP 41, RFC [RFC2914] Floyd, S., "Congestion Control Principles", BCP 41,
2914, September 2000. RFC 2914, DOI 10.17487/RFC2914, September 2000,
<http://www.rfc-editor.org/info/rfc2914>.
[RFC4340] Kohler, E., Handley, M., and S. Floyd, "Datagram [RFC4340] Kohler, E., Handley, M., and S. Floyd, "Datagram
Congestion Control Protocol (DCCP)", RFC 4340, March 2006. Congestion Control Protocol (DCCP)", RFC 4340,
DOI 10.17487/RFC4340, March 2006,
<http://www.rfc-editor.org/info/rfc4340>.
[RFC4960] Stewart, R., "Stream Control Transmission Protocol", RFC [RFC4960] Stewart, R., Ed., "Stream Control Transmission Protocol",
4960, September 2007. RFC 4960, DOI 10.17487/RFC4960, September 2007,
<http://www.rfc-editor.org/info/rfc4960>.
[RFC5348] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP [RFC5348] Floyd, S., Handley, M., Padhye, J., and J. Widmer, "TCP
Friendly Rate Control (TFRC): Protocol Specification", RFC Friendly Rate Control (TFRC): Protocol Specification",
5348, September 2008. RFC 5348, DOI 10.17487/RFC5348, September 2008,
<http://www.rfc-editor.org/info/rfc5348>.
[RFC5559] Eardley, P., "Pre-Congestion Notification (PCN) [RFC5559] Eardley, P., Ed., "Pre-Congestion Notification (PCN)
Architecture", RFC 5559, June 2009. Architecture", RFC 5559, DOI 10.17487/RFC5559, June 2009,
<http://www.rfc-editor.org/info/rfc5559>.
[RFC6057] Bastian, C., Klieber, T., Livingood, J., Mills, J., and R. [RFC6057] Bastian, C., Klieber, T., Livingood, J., Mills, J., and R.
Woundy, "Comcast's Protocol-Agnostic Congestion Management Woundy, "Comcast's Protocol-Agnostic Congestion Management
System", RFC 6057, December 2010. System", RFC 6057, DOI 10.17487/RFC6057, December 2010,
<http://www.rfc-editor.org/info/rfc6057>.
[RFC6789] Briscoe, B., Woundy, R., and A. Cooper, "Congestion [RFC6789] Briscoe, B., Ed., Woundy, R., Ed., and A. Cooper, Ed.,
Exposure (ConEx) Concepts and Use Cases", RFC 6789, "Congestion Exposure (ConEx) Concepts and Use Cases",
December 2012. RFC 6789, DOI 10.17487/RFC6789, December 2012,
<http://www.rfc-editor.org/info/rfc6789>.
[RFC6817] Shalunov, S., Hazel, G., Iyengar, J., and M. Kuehlewind, [RFC6817] Shalunov, S., Hazel, G., Iyengar, J., and M. Kuehlewind,
"Low Extra Delay Background Transport (LEDBAT)", RFC 6817, "Low Extra Delay Background Transport (LEDBAT)", RFC 6817,
December 2012. DOI 10.17487/RFC6817, December 2012,
<http://www.rfc-editor.org/info/rfc6817>.
[RFC7414] Duke, M., Braden, R., Eddy, W., Blanton, E., and A. [RFC7414] Duke, M., Braden, R., Eddy, W., Blanton, E., and A.
Zimmermann, "A Roadmap for Transmission Control Protocol Zimmermann, "A Roadmap for Transmission Control Protocol
(TCP) Specification Documents", RFC 7414, February 2015. (TCP) Specification Documents", RFC 7414,
DOI 10.17487/RFC7414, February 2015,
[Ren12] Ren, Y., Zhao, Y., and P. Liu, "A survey on TCP Incast in <http://www.rfc-editor.org/info/rfc7414>.
data center networks, International Journal of
Communication Systems, Volume 27, Issue 8, pages
1160-117", 1990.
[Shr96] Shreedhar, M. and G. Varghese, "Efficient Fair Queueing [Shr96] Shreedhar, M. and G. Varghese, "Efficient Fair Queueing
Using Deficit Round Robin", IEEE/ACM Transactions on Using Deficit Round Robin", IEEE/ACM Transactions on
Networking Vol 4, No. 3 , July 1996. Networking, Vol. 4, No. 3, July 1996.
[Sto97] Stoica, I. and H. Zhang, "A Hierarchical Fair Service [Sto97] Stoica, I. and H. Zhang, "A Hierarchical Fair Service
Curve algorithm for Link sharing, real-time and priority Curve algorithm for Link sharing, real-time and priority
services", ACM SIGCOMM , 1997. services", ACM SIGCOMM, 1997.
[Sut99] Suter, B., "Buffer Management Schemes for Supporting TCP [Sut99] Suter, B., "Buffer Management Schemes for Supporting TCP
in Gigabit Routers with Per-flow Queueing", IEEE Journal in Gigabit Routers with Per-flow Queueing", IEEE Journal
on Selected Areas in Communications Vol. 17 Issue 6, June, on Selected Areas in Communications, Vol. 17, Issue 6, pp.
1999, pp. 1159-1169. , 1999. 1159-1169, June 1999.
[Willinger95] [Willinger95]
Willinger, W., Taqqu, M., Sherman, R., Wilson, D., and V. Willinger, W., Taqqu, M., Sherman, R., Wilson, D., and V.
Jacobson, "Self-Similarity Through High-Variability: Jacobson, "Self-Similarity Through High-Variability:
Statistical Analysis of Ethernet LAN Traffic at the Source Statistical Analysis of Ethernet LAN Traffic at the Source
Level", SIGCOMM Symposium proceedings on Communications Level", SIGCOMM Symposium proceedings on Communications
architectures and protocols , August 1995. architectures and protocols, August 1995.
[Zha90] Zhang, L. and D. Clark, "Oscillating Behavior of Network [Zha90] Zhang, L. and D. Clark, "Oscillating Behavior of Network
Traffic: A Case Study Simulation, Traffic: A Case Study Simulation", 1990,
http://groups.csail.mit.edu/ana/Publications/Zhang-DDC- <http://groups.csail.mit.edu/ana/Publications/Zhang-DDC-
Oscillating-Behavior-of-Network-Traffic-1990.pdf", 1990. Oscillating-Behavior-of-Network-Traffic-1990.pdf>.
Appendix A. Change Log
RFC-Editor please remove this appendix before publication.
Initial Version: March 2013
Minor update of the algorithms that the IETF recommends SHOULD NOT
require operational (especially manual) configuration or tuning
April 2013
Major surgery. This draft is for discussion at IETF-87 and expected
to be further updated.
July 2013
-00 WG Draft - Updated transport recommendations; revised deployment
configuration section; numerous minor edits.
Oct 2013
-01 WG Draft - Updated transport recommendations; revised deployment
configuration section; numerous minor edits.
Jan 2014 - Feedback from WG.
-02 WG Draft - Minor edits Feb 2014 - Mainly language fixes.
-03 WG Draft - Minor edits Feb 2013 - Comments from David Collier-
Brown and David Taht.
-04 WG Draft - Minor edits May 2014 - Comments during WGLC: Provided
some introductory subsections to help people (with subsections and
better text). - Written more on the role scheduling. - Clarified
that ECN mark threshold needs to be configurable. - Reworked your
"knee" para. Various updates in response to feedback.
-05 WG Draft - Minor edits June 2014 - New text added to address
further comments, and improve introduction - adding context,
reference to Conex, linking between sections, added text on
synchronization.
-06 WG Draft - Minor edits July 2014 - Reorganised the introduction
following WG feedback to better explain how this relates to the
original goals of RFC2309. Added item on packet bursts. Various
minor corrections incorporated - no change to main
recommendations.
-07 WG Draft - Minor edits July 2014 - Replaced ID REF by RFC 7141.
Changes made to introduction following inputs from Wes Eddy and
John Leslie. Corrections and additions proposed by Bob Briscoe.
-08 WG Draft - Minor edits August 2014 - Review comments from John Acknowledgements
Leslie and Bob Briscoe. Text corrections including; updated
Acknowledgments (RFC2309 ref) s/congestive/congestion/g; changed
the more bold language from RFC2309 to reflect a more considered
perceived threat to Internet Performance; modified the category
that is not-TCP-like to be "less responsive to congestion than
TCP" and more clearkly noted that represents a range of
behaviours.
-09 WG Draft - Minor edits Jan 2015 - Edits following LC comments. The original draft of this document describing best current practice
was based on [RFC2309], an Informational RFC. It was written by the
End-to-End Research Group, which is to say Bob Braden, Dave Clark,
Jon Crowcroft, Bruce Davie, Steve Deering, Deborah Estrin, Sally
Floyd, Van Jacobson, Greg Minshall, Craig Partridge, Larry Peterson,
KK Ramakrishnan, Scott Shenker, John Wroclawski, and Lixia Zhang.
Although there are important differences, many of the key arguments
in the present document remain unchanged from those in RFC 2309.
-10 WG Draft - Minor edits Feb 2015 - Update following IESG Review The need for an updated document was agreed to in the TSV area
meeting at IETF 86. This document was reviewed on the aqm@ietf.org
list. Comments were received from Colin Perkins, Richard
Scheffenegger, Dave Taht, John Leslie, David Collier-Brown, and many
others.
-11 WG Draft - Minor edits Feb 2015 - Resolution of last issues. Gorry Fairhurst was in part supported by the European Community under
its Seventh Framework Programme through the Reducing Internet
Transport Latency (RITE) project (ICT-317700).
Authors' Addresses Authors' Addresses
Fred Baker (editor) Fred Baker (editor)
Cisco Systems Cisco Systems
Santa Barbara, California 93117 Santa Barbara, California 93117
USA United States
Email: fred@cisco.com Email: fred@cisco.com
Godred Fairhurst (editor) Godred Fairhurst (editor)
University of Aberdeen University of Aberdeen
School of Engineering School of Engineering
Fraser Noble Building Fraser Noble Building
Aberdeen, Scotland AB24 3UE Aberdeen, Scotland AB24 3UE
UK United Kingdom
Email: gorry@erg.abdn.ac.uk Email: gorry@erg.abdn.ac.uk
URI: http://www.erg.abdn.ac.uk URI: http://www.erg.abdn.ac.uk
 End of changes. 202 change blocks. 
553 lines changed or deleted 525 lines changed or added

This html diff was produced by rfcdiff 1.42. The latest version is available from http://tools.ietf.org/tools/rfcdiff/