draft-ietf-tcpm-cubic-06.txt   draft-ietf-tcpm-cubic-07.txt 
TCP Maintenance and Minor Extensions (TCPM) WG I. Rhee TCP Maintenance and Minor Extensions (TCPM) WG I. Rhee
Internet-Draft NCSU Internet-Draft NCSU
Intended status: Informational L. Xu Intended status: Informational L. Xu
Expires: March 21, 2018 UNL Expires: May 17, 2018 UNL
S. Ha S. Ha
Colorado Colorado
A. Zimmermann A. Zimmermann
L. Eggert L. Eggert
NetApp NetApp
R. Scheffenegger R. Scheffenegger
September 17, 2017 November 13, 2017
CUBIC for Fast Long-Distance Networks CUBIC for Fast Long-Distance Networks
draft-ietf-tcpm-cubic-06 draft-ietf-tcpm-cubic-07
Abstract Abstract
CUBIC is an extension to the current TCP standards. The protocol CUBIC is an extension to the current TCP standards. It differs from
differs from the current TCP standards only in the congestion window the current TCP standards only in the congestion control algorithm in
adjustment function in the sender side. In particular, it uses a the sender side. In particular, it uses a cubic function instead of
cubic function instead of a linear window increase function of the a linear window increase function of the current TCP standards to
current TCP standards to improve scalability and stability under fast improve scalability and stability under fast and long distance
and long distance networks. CUBIC and its predecessor algorithm have networks. CUBIC and its predecessor algorithm have been adopted as
been adopted as default by Linux and have been used for many years. default by Linux and have been used for many years. This document
This document provides a specification of CUBIC to enable third party provides a specification of CUBIC to enable third party
implementation and to solicit the community feedback through implementations and to solicit the community feedback through
experimentation on the performance of CUBIC. experimentation on the performance of CUBIC.
Status of This Memo Status of This Memo
This Internet-Draft is submitted in full conformance with the This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79. provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet- working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/. Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress." material or to cite them other than as "work in progress."
This Internet-Draft will expire on March 21, 2018. This Internet-Draft will expire on May 17, 2018.
Copyright Notice Copyright Notice
Copyright (c) 2017 IETF Trust and the persons identified as the Copyright (c) 2017 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(https://trustee.ietf.org/license-info) in effect on the date of (https://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License. described in the Simplified BSD License.
Table of Contents Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Conventions . . . . . . . . . . . . . . . . . . . . . . . . . 3 2. Conventions . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Design principle of CUBIC . . . . . . . . . . . . . . . . . . 4 3. Design principles of CUBIC . . . . . . . . . . . . . . . . . 4
4. CUBIC Congestion Control . . . . . . . . . . . . . . . . . . 5 4. CUBIC Congestion Control . . . . . . . . . . . . . . . . . . 6
4.1. Window growth function . . . . . . . . . . . . . . . . . 6 4.1. Window increase function . . . . . . . . . . . . . . . . 6
4.2. TCP-friendly region . . . . . . . . . . . . . . . . . . . 7 4.2. TCP-friendly region . . . . . . . . . . . . . . . . . . . 7
4.3. Concave region . . . . . . . . . . . . . . . . . . . . . 7 4.3. Concave region . . . . . . . . . . . . . . . . . . . . . 8
4.4. Convex region . . . . . . . . . . . . . . . . . . . . . . 7 4.4. Convex region . . . . . . . . . . . . . . . . . . . . . . 8
4.5. Multiplicative decrease . . . . . . . . . . . . . . . . . 8 4.5. Multiplicative decrease . . . . . . . . . . . . . . . . . 8
4.6. Fast convergence . . . . . . . . . . . . . . . . . . . . 8 4.6. Fast convergence . . . . . . . . . . . . . . . . . . . . 9
4.7. Timeout . . . . . . . . . . . . . . . . . . . . . . . . . 9 4.7. Timeout . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.8. Slowstart . . . . . . . . . . . . . . . . . . . . . . . . 9 4.8. Slowstart . . . . . . . . . . . . . . . . . . . . . . . . 10
5. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 9 5. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 10
5.1. Fairness to standard TCP . . . . . . . . . . . . . . . . 10 5.1. Fairness to standard TCP . . . . . . . . . . . . . . . . 10
5.2. Using Spare Capacity . . . . . . . . . . . . . . . . . . 12 5.2. Using Spare Capacity . . . . . . . . . . . . . . . . . . 12
5.3. Difficult Environments . . . . . . . . . . . . . . . . . 12 5.3. Difficult Environments . . . . . . . . . . . . . . . . . 13
5.4. Investigating a Range of Environments . . . . . . . . . . 13 5.4. Investigating a Range of Environments . . . . . . . . . . 13
5.5. Protection against Congestion Collapse . . . . . . . . . 13 5.5. Protection against Congestion Collapse . . . . . . . . . 13
5.6. Fairness within the Alternative Congestion Control 5.6. Fairness within the Alternative Congestion Control
Algorithm. . . . . . . . . . . . . . . . . . . . . . . . 13 Algorithm. . . . . . . . . . . . . . . . . . . . . . . . 13
5.7. Performance with Misbehaving Nodes and Outside Attackers 13 5.7. Performance with Misbehaving Nodes and Outside Attackers 13
5.8. Behavior for Application-Limited Flows . . . . . . . . . 13 5.8. Behavior for Application-Limited Flows . . . . . . . . . 13
5.9. Responses to Sudden or Transient Events . . . . . . . . . 14 5.9. Responses to Sudden or Transient Events . . . . . . . . . 14
5.10. Incremental Deployment . . . . . . . . . . . . . . . . . 14 5.10. Incremental Deployment . . . . . . . . . . . . . . . . . 14
6. Security Considerations . . . . . . . . . . . . . . . . . . . 14 6. Security Considerations . . . . . . . . . . . . . . . . . . . 14
7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 14 7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 14
8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 14 8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 14
9. References . . . . . . . . . . . . . . . . . . . . . . . . . 14 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 14
9.1. Normative References . . . . . . . . . . . . . . . . . . 14 9.1. Normative References . . . . . . . . . . . . . . . . . . 14
9.2. Informative References . . . . . . . . . . . . . . . . . 15 9.2. Informative References . . . . . . . . . . . . . . . . . 15
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 16 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 16
1. Introduction 1. Introduction
The low utilization problem of TCP in fast long-distance networks is The low utilization problem of TCP in fast long-distance networks is
well documented in [K03][RFC3649]. This problem arises from a slow well documented in [K03] [RFC3649]. This problem arises from a slow
increase of congestion window following a congestion event in a increase of congestion window following a congestion event in a
network with a large bandwidth delay product (BDP). Our experience network with a large bandwidth delay product (BDP). Experience
[HKLRX06] indicates that this problem is frequently observed even in [HKLRX06] indicates that this problem is frequently observed even in
the range of congestion window sizes over several hundreds of packets the range of congestion window sizes over several hundreds of packets
(each packet is sized around 1000 bytes) especially under a network especially under a network path with over 100ms round-trip times
path with over 100ms round-trip times (RTTs). This problem is (RTTs). This problem is equally applicable to all Reno style TCP
equally applicable to all Reno style TCP standards and their standards and their variants, including TCP-RENO [RFC5681], TCP-
variants, including TCP-RENO [RFC5681], TCP-NewReno NewReno [RFC6582] [RFC6675], SCTP [RFC4960], TFRC [RFC5348] that use
[RFC6582][RFC6675], SCTP [RFC4960], TFRC [RFC5348] that use the same the same linear increase function for window growth, which we refer
linear increase function for window growth, which we refer to to collectively as Standard TCP below.
collectively as Standard TCP below.
CUBIC [HRX08] is a modification to the congestion control mechanism CUBIC, originally proposed in [HRX08], is a modification to the
of Standard TCP, in particular, to the window increase function of congestion control algorithm of Standard TCP to remedy this problem.
Standard TCP senders, to remedy this problem. Specifically, it uses This document describes the most recent specification of CUBIC.
a cubic function instead of a linear window increase function of the Specifically, CUBIC uses a cubic function instead of a linear window
Standrad TCP to improve scalability and stability under fast and long increase function of Standard TCP to improve scalability and
distance networks. stability under fast and long distance networks.
BIC-TCP, a predecessor of CUBIC, has been selected as default TCP BIC-TCP [XHR04], a predecessor of CUBIC, has been selected as the
congestion control algorithm by Linux in the year 2005 and been used default TCP congestion control algorithm by Linux in the year 2005
for several years by the Internet community at large. CUBIC uses a and been used for several years by the Internet community at large.
similar window growth function as BIC-TCP and is designed to be less CUBIC uses a similar window increase function as BIC-TCP and is
aggressive and fairer to TCP in bandwidth usage than BIC-TCP while designed to be less aggressive and fairer to Standard TCP in
maintaining the strengths of BIC-TCP such as stability, window bandwidth usage than BIC-TCP while maintaining the strengths of BIC-
scalability and RTT fairness. CUBIC has already been deployed TCP such as stability, window scalability and RTT fairness. CUBIC
globally by Linux. Through extensive testing in various Internet has already replaced BIC-TCP as the default TCP congestion control
scenarios, we believe that CUBIC is safe for testing and deployment algorithm in Linux and has been deployed globally by Linux. Through
in the global Internet. extensive testing in various Internet scenarios, we believe that
CUBIC is safe for testing and deployment in the global Internet.
In the ensuing sections, we first brefly explain the design principle In the following sections, we first briefly explain the design
of CUBIC, then provide the exact specification of CUBIC, and finally principles of CUBIC, then provide the exact specification of CUBIC,
discuss the safety features of CUBIC following the guidelines and finally discuss the safety features of CUBIC following the
specified in [RFC5033]. guidelines specified in [RFC5033].
2. Conventions 2. Conventions
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].
3. Design principle of CUBIC 3. Design principles of CUBIC
CUBIC [HRX08] uses a cubic window increase function in terms of the CUBIC is designed according to the following design principles.
elapsed time from the last congestion event. While most alternative
algorithms to Standard TCP uses a convex increase function where Principle 1: For better network utilization and stability, CUBIC
during congestion avoidance the window increment is always uses both the concave and convex profiles of a cubic function to
increasing, CUBIC uses both the concave and convex profiles of a increase the congestion window size, instead of using just a
cubic function for window increase. After a window reduction convex function.
following a loss event detected by duplicate ACKs, it registers the
window size where it got the loss event as W_max and performs a Principle 2: To be TCP-friendly, CUBIC is designed to behave like
multiplicative decrease of congestion window and the regular fast Standard TCP in networks with short RTTs and small bandwidth where
recovery and retransmit of Standard TCP. After it enters into Standard TCP performs well.
congestion avoidance from fast recovery, it starts to increase the
window using the concave profile of the cubic function. The cubic Principle 3: For RTT-fairness, CUBIC is designed to achieve linear
function is set to have its plateau at W_max so the concave growth bandwidth share among flows with different RTTs.
continues until the window size becomes W_max. After that, the cubic
function turns into a convex profile and the convex window growth Principle 4: CUBIC appropriately sets its multiplicative window
begins. This style of window adjustment (concave and then convex) decrease factor, in order to balance between the scalability and
improves protocol and network stability while maintaining high convergence speed.
Principle 1: For better network utilization and stability, CUBIC
[HRX08] uses a cubic window increase function in terms of the elapsed
time from the last congestion event. While most alternative
congestion control algorithms to Standard TCP increase the congestion
window using convex functions, CUBIC uses both the concave and convex
profiles of a cubic function for window growth. After a window
reduction in response to a congestion event detected by duplicate
ACKs or ECN-Echo ACKs[RFC3168], CUBIC registers the congestion window
size where it got the congestion event as W_max and performs a
multiplicative decrease of congestion window. After it enters into
congestion avoidance, it starts to increase the congestion window
using the concave profile of the cubic function. The cubic function
is set to have its plateau at W_max so that the concave window
increase continues until the window size becomes W_max. After that,
the cubic function turns into a convex profile and the convex window
increase begins. This style of window adjustment (concave and then
convex) improves the algorithm stability while maintaining high
network utilization [CEHRX07]. This is because the window size network utilization [CEHRX07]. This is because the window size
remains almost constant, forming a plateau around W_max where network remains almost constant, forming a plateau around W_max where network
utilization is deemed highest and under steady state, most window utilization is deemed highest. Under steady state, most window size
size samples of CUBIC are close to W_max, thus promoting high network samples of CUBIC are close to W_max, thus promoting high network
utilization and protocol stability. Note that protocols with convex utilization and stability. Note that those congestion control
increase functions have the maximum increments around W_max and algorithms using only convex functions to increase the congestion
introduces a large number of packet bursts around the saturation window size have the maximum increments around W_max and thus
point of the network, likely causing frequent global loss introduce a large number of packet bursts around the saturation point
synchronizations. of the network, likely causing frequent global loss synchronizations.
Another notable feature of CUBIC is that its window increase rate is Principle 2: CUBIC promotes per-flow fairness to Standard TCP. Note
mostly independent of RTT, and follows a (cubic) function of the that Standard TCP performs well under short RTT and small bandwidth
elapsed time from the beginning of congestion avoidance. This (or small BDP) networks. Only in long RTT and large bandwidth (or
feature promotes per-flow fairness to Standard TCP as well as RTT- large BDP) networks, it has the scalability problem. An alternative
fairness. Note that Standard TCP performs well under short RTT and congestion control algorithm to Standard TCP designed to be friendly
small bandwidth (or small BDP) networks. Only in a large long RTT to Standard TCP at a per-flow basis must operate to increase its
and large bandwidth (or large BDP) networks, it has the scalability congestion window less aggressively in small BDP networks than in
problem. An alternative protocol to Standard TCP designed to be large BDP networks. The aggressiveness of CUBIC mainly depends on
friendly to Standard TCP at a per-flow basis must operate to increase the maximum window size before a window reduction, which is smaller
its window much less aggressively in small BDP networks than in large in small BDP networks than in large BDP networks. Thus, CUBIC
BDP networks. In CUBIC, its window growth rate is slowest around the increases its congestion window less aggressively in small BDP
inflection point of the cubic function and this function does not networks than in large BDP networks. Furthermore, in cases when the
depend on RTT. In a smaller BDP network where Standard TCP flows are cubic function of CUBIC increases its congestion window less
working well, the absolute amount of the window decrease at a loss aggressively than Standard TCP, CUBIC simply follows the window size
event is always smaller because of the multiplicative decrease. of Standard TCP to ensure that CUBIC achieves at least the same
Therefore, in CUBIC, the starting window size after a loss event from throughput as Standard TCP in small BDP networks. We call this
which the window starts to increase, is smaller in a smaller BDP region where CUBIC behaves like Standard TCP, the TCP-friendly
network, thus falling nearer to the plateau of the cubic function region.
where the growth rate is slowest. By setting appropriate values of
the cubic function parameters, CUBIC sets its growth rate always no
faster than Standard TCP around its inflection point. When the cubic
function grows slower than the window of Standard TCP, CUBIC simply
follows the window size of Standard TCP to ensure fairness to
Standard TCP in a small BDP network. We call this region where CUBIC
behaves like Standard TCP, the TCP-friendly region.
CUBIC maintains the same window growth rate independent of RTTs Principle 3: Two CUBIC flows with different RTTs have their
outside of the TCP-friendly region, and flows with different RTTs throughput ratio linearly proportional to the inverse of their RTT
have the similar window sizes under steady state when they operate ratio, where the throughput of a flow is approximately its congestion
outside the TCP-friendly region. This ensures CUBIC flows with window size divided by its RTT. Specifically, CUBIC maintains a
different RTTs to have their bandwidth shares (approximately, window/ window increase rate independent of RTTs outside of the TCP-friendly
RTT) linearly proportional to the inverse of their RTT ratio (the region, and thus flows with different RTTs have similar congestion
longer RTT, the smaller the share). This behavior is the same as window sizes under steady state when they operate outside the TCP-
that of Standard TCP under high statistical multiplexing environments friendly region. This notion of a linear throughput ratio is similar
where packet losses are independent of individual flow rates. to that of Standard TCP under high statistical multiplexing
However, under low statistical multiplexing environments, the environments where packet losses are independent of individual flow
bandwidth share ratio of Standard TCP flows with different RTTs is rates. However, under low statistical multiplexing environments, the
squarely proportional to the inverse of their RTT ratio [XHR04]. throughput ratio of Standard TCP flows with different RTTs is
CUBIC always ensures the linear ratio independent of the levels of quadratically proportional to the inverse of their RTT ratio [XHR04].
statistical multiplexing. This is an improvement over Standard TCP. CUBIC always ensures the linear throughput ratio independent of the
While there is no consensus on a particular bandwidth share ratios of levels of statistical multiplexing. This is an improvement over
different RTT flows, we believe that under wired Internet, use of the Standard TCP. While there is no consensus on particular throughput
linear share notion seems more reasonable than equal share or a ratios of different RTT flows, we believe that under wired Internet,
higher order shares. HTCP [LS08] currently uses the equal share. use of a linear throughput ratio seems more reasonable than equal
throughputs (i.e., same throughput for flows with different RTTs) or
a higher order throughput ratio (e.g., a quadratical throughput ratio
of Standard TCP under low statistical multiplexing environments).
CUBIC sets the multiplicative window decrease factor to 0.7 while Principle 4: To balance between the scalability and convergence
Standard TCP uses 0.5. While this improves the scalability of the speed, CUBIC sets the multiplicative window decrease factor to 0.7
protocol, a side effect of this decision is slower convergence while Standard TCP uses 0.5. While this improves the scalability of
CUBIC, a side effect of this decision is slower convergence
especially under low statistical multiplexing environments. This especially under low statistical multiplexing environments. This
design choice is following the observation that the author of HSTCP design choice is following the observation that the author of HSTCP
[RFC3649] has made along with other researchers (e.g., [GV02]): the [RFC3649] has made along with other researchers (e.g., [GV02]): the
current Internet becomes more asynchronous with less frequent loss current Internet becomes more asynchronous with less frequent loss
synchronizations with high statistical multiplexing. Under this synchronizations with high statistical multiplexing. Under this
environment, even strict Multiplicative-Increase Multiplicative- environment, even strict Multiplicative-Increase Multiplicative-
Decrease (MIMD) can converge. CUBIC flows with the same RTT always Decrease (MIMD) can converge. CUBIC flows with the same RTT always
converge to the same share of bandwidth independent of statistical converge to the same throughput independent of statistical
multiplexing, thus achieving intra-protocol fairness. We also find multiplexing, thus achieving intra-algorithm fairness. We also find
that under the environments with sufficient statistical multiplexing, that under the environments with sufficient statistical multiplexing,
the convergence speed of CUBIC flows is reasonable. the convergence speed of CUBIC flows is reasonable.
4. CUBIC Congestion Control 4. CUBIC Congestion Control
The unit of all window sizes in this document is segments of the The unit of all window sizes in this document is segments of the
maximum segment size (MSS), and the unit of all times is seconds. maximum segment size (MSS), and the unit of all times is seconds.
Let cwnd denote the congestion window size of a flow, and ssthresh
denote the slow start threshold.
4.1. Window growth function 4.1. Window increase function
CUBIC maintains the acknowledgment (ACK) clocking of Standard TCP by CUBIC maintains the acknowledgment (ACK) clocking of Standard TCP by
increasing congestion window only at the reception of ACK. The increasing congestion window only at the reception of ACK. It does
protocol does not make any change to the fast recovery and retransmit not make any change to the fast recovery and retransmit of TCP, such
of TCP, such as TCP-NewReno [RFC6582] [RFC6675]. During congestion as TCP-NewReno [RFC6582] [RFC6675]. During congestion avoidance
avoidance after fast recovery, CUBIC changes the window update after a congestion event where a packet loss is detected by duplicate
algorithm of Standard TCP. Suppose that W_max is the window size ACKs or a network congestion is detected by ACKs with ECN-Echo flags
before the window is reduced in the last fast retransmit and [RFC3168], CUBIC changes the window increase function of Standard
recovery. TCP. Suppose that W_max is the window size just before the window is
reduced in the last congestion event.
The window growth function of CUBIC uses the following function: CUBIC uses the following window increase function:
W_cubic(t) = C*(t-K)^3 + W_max (Eq. 1) W_cubic(t) = C*(t-K)^3 + W_max (Eq. 1)
where C is a constant fixed to determine the aggressiveness of window where C is a constant fixed to determine the aggressiveness of window
growth in high BDP networks, t is the elapsed time from the last increase in high BDP networks, t is the elapsed time from the
window reduction that is measured right after the fast recovery in beginning of the current congestion avoidance, and K is the time
response to duplicate ACKs or after the congestion window reduction period that the above function takes to increase the current window
in response to ECN-Echo ACKs, and K is the time period that the above size to W_max if there are no further congestion events and is
function takes to increase the current window size to W_max if there calculated using the following equation:
is no further loss event and is calculated by using the following
equation:
K = cubic_root(W_max*(1-beta_cubic)/C) (Eq. 2) K = cubic_root(W_max*(1-beta_cubic)/C) (Eq. 2)
where beta_cubic is the CUBIC multiplication decrease factor, that where beta_cubic is the CUBIC multiplication decrease factor, that
is, when a packet loss detected by duplicate ACKs or a network is, when a congestion event is detected, CUBIC reduces its cwnd to
congestion detected by ECN-Echo ACKs occurs, CUBIC reduces its W_cubic(0)=W_max*beta_cubic. We discuss how we set beta_cubic in
current window cwnd to W_cubic(0)=W_max*beta_cubic. We discuss how Section 4.5 and how we set C in Section 5.
we set beta_cubic in Section 4.5 and how we set C in Section 5.
Upon receiving an ACK during congestion avoidance, CUBIC computes the Upon receiving an ACK during congestion avoidance, CUBIC computes the
window growth rate during the next RTT period using Eq. 1. It sets window increase rate during the next RTT period using Eq. 1. It sets
W_cubic(t+RTT) as the candidate target value of congestion window, W_cubic(t+RTT) as the candidate target value of congestion window,
where RTT is the weithed average RTT calculated by the standard TCP. where RTT is the weighted average RTT calculated by Standard TCP.
Depending on the value of the current window size cwnd, CUBIC runs in Depending on the value of the current congestion window size cwnd,
three different modes. CUBIC runs in three different modes.
1) The TCP-friendly region, which ensures that CUBIC achieves at 1) The TCP-friendly region, which ensures that CUBIC achieves at
least the same throughput as the standard TCP. least the same throughput as Standard TCP.
2) The concave region, if CUBIC is not in the TCP-friendly region 2) The concave region, if CUBIC is not in the TCP-friendly region
and cwnd is less than W_max. and cwnd is less than W_max.
3) The convex region, if CUBIC is not in the TCP-friendly region 3) The convex region, if CUBIC is not in the TCP-friendly region
and cwnd is greater than W_max. and cwnd is greater than W_max.
Below, we describe the exact actions taken by CUBIC in each region. Below, we describe the exact actions taken by CUBIC in each region.
4.2. TCP-friendly region 4.2. TCP-friendly region
Standard TCP performs well in certain types of networks, for example, Standard TCP performs well in certain types of networks, for example,
under short RTT and small bandwidth (or small BDP) networks. In under short RTT and small bandwidth (or small BDP) networks. In
these networks, we use the TCP-friendly region to ensure that CUBIC these networks, we use the TCP-friendly region to ensure that CUBIC
achieves at least the same throughput as the standard TCP. achieves at least the same throughput as Standard TCP.
The TCP-friendly region is designed according to the analysis The TCP-friendly region is designed according to the analysis
described in [FHP00]. The analysis studies the performance of an described in [FHP00]. The analysis studies the performance of an
Additive Increase and Multiplicative Decrease (AIMD) algorithm with Additive Increase and Multiplicative Decrease (AIMD) algorithm with
an additive factor of alpha_aimd (segment per RTT) and a an additive factor of alpha_aimd (segments per RTT) and a
multiplicative factor of beta_aimd, denoted by AIMD(alpha_aimd, multiplicative factor of beta_aimd, denoted by AIMD(alpha_aimd,
beta_aimd). Specifically, the average window size of beta_aimd). Specifically, the average congestion window size of
AIMD(alpha_aimd, beta_aimd) can be calculated using Eq. 3. The AIMD(alpha_aimd, beta_aimd) can be calculated using Eq. 3. The
analysis shows that AIMD(alpha_aimd, beta_aimd) with analysis shows that AIMD(alpha_aimd, beta_aimd) with
alpha_aimd=3*(1-beta_aimd)/(1+beta_aimd) achieves the same average alpha_aimd=3*(1-beta_aimd)/(1+beta_aimd) achieves the same average
window size as the standard TCP that uses AIMD(1, 0.5). window size as Standard TCP that uses AIMD(1, 0.5).
AVG_W_aimd = [ alpha_aimd * (1+beta_aimd) / AVG_W_aimd = [ alpha_aimd * (1+beta_aimd) /
(2*(1-beta_aimd)*p) ]^0.5 (Eq. 3) (2*(1-beta_aimd)*p) ]^0.5 (Eq. 3)
Based on the above analysis, CUBIC uses Eq. 4 to estimate the window Based on the above analysis, CUBIC uses Eq. 4 to estimate the window
size W_est of AIMD(alpha_aimd, beta_aimd) with size W_est of AIMD(alpha_aimd, beta_aimd) with
alpha_aimd=3*(1-beta_cubic)/(1+beta_cubic) and beta_aimd=beta_cubic, alpha_aimd=3*(1-beta_cubic)/(1+beta_cubic) and beta_aimd=beta_cubic,
which achieves the same average window size as the standard TCP. which achieves the same average window size as Standard TCP. When
When receiving an ACK in congestion avoidance (cwnd could be greater receiving an ACK in congestion avoidance (cwnd could be greater than
than or less than W_max), CUBIC checks whether W_cubic(t) is less or less than W_max), CUBIC checks whether W_cubic(t) is less than
than W_est(t). If so, CUBIC is in the TCP-friendly region and cwnd W_est(t). If so, CUBIC is in the TCP-friendly region and cwnd SHOULD
SHOULD be set to W_est(t) at each reception of ACK. be set to W_est(t) at each reception of ACK.
W_est(t) = W_max*beta_cubic + W_est(t) = W_max*beta_cubic +
[3*(1-beta_cubic)/(1+beta_cubic)] * (t/RTT) (Eq. 4) [3*(1-beta_cubic)/(1+beta_cubic)] * (t/RTT) (Eq. 4)
4.3. Concave region 4.3. Concave region
When receiving an ACK in congestion avoidance, if the protocol is not When receiving an ACK in congestion avoidance, if CUBIC is not in the
in the TCP-friendly region and cwnd is less than W_max, then the TCP-friendly region and cwnd is less than W_max, then CUBIC is in the
protocol is in the concave region. In this region, cwnd MUST be concave region. In this region, cwnd MUST be incremented by
incremented by (W_cubic(t+RTT) - cwnd)/cwnd for each received ACK, (W_cubic(t+RTT) - cwnd)/cwnd for each received ACK, where
where W_cubic(t+RTT) is calculated using Eq. 1. W_cubic(t+RTT) is calculated using Eq. 1.
4.4. Convex region 4.4. Convex region
When the current window size of CUBIC is larger than W_max, it passes When receiving an ACK in congestion avoidance, if CUBIC is not in the
the plateau of the cubic function after which CUBIC follows the TCP-friendly region and cwnd is larger than or equal to W_max, then
convex profile of the cubic function. Since cwnd is larger than the CUBIC is in the convex region. The convex region indicates that the
previous saturation point W_max, this indicates that the network network conditions might have been perturbed since the last
conditions might have been perturbed since the last loss event, congestion event, possibly implying more available bandwidth after
possibly implying more available bandwidth after some flow some flow departures. Since the Internet is highly asynchronous,
departures. Since the Internet is highly asynchronous, some amount some amount of perturbation is always possible without causing a
of perturbation is always possible without causing a major change in major change in available bandwidth. In this region, CUBIC is being
available bandwidth. In this phase, CUBIC is being very careful by very careful by very slowly increasing its window size. The convex
very slowly increasing its window size. The convex profile ensures profile ensures that the window increases very slowly at the
that the window increases very slowly at the beginning and gradually beginning and gradually increases its increase rate. We also call
increases its growth rate. We also call this phase as the maximum this region as the maximum probing phase since CUBIC is searching for
probing phase since CUBIC is searching for a new W_max. In this a new W_max. In this region, cwnd MUST be incremented by
region, cwnd MUST be incremented by (W_cubic(t+RTT) - cwnd)/cwnd for (W_cubic(t+RTT) - cwnd)/cwnd for each received ACK, where
each received ACK, where W_cubic(t+RTT) is calculated using Eq. 1. W_cubic(t+RTT) is calculated using Eq. 1.
4.5. Multiplicative decrease 4.5. Multiplicative decrease
When a packet loss detected by duplicate ACKs or a network congestion When a packet loss is detected by duplicate ACKs or a network
detected by ECN-Echo ACKs occurs, CUBIC updates its W_max, cwnd, and congestion is detected by ECN-Echo ACKs, CUBIC updates its W_max,
ssthresh (slow start threshold) as follows. Parameter beta_cubic cwnd, and ssthresh (slow start threshold) as follows. Parameter
SHOULD be set to 0.7. beta_cubic SHOULD be set to 0.7.
W_max = cwnd; // save window size before reduction W_max = cwnd; // save window size before reduction
ssthresh = cwnd * beta_cubic; // new slow start threshold ssthresh = cwnd * beta_cubic; // new slow start threshold
ssthresh = max(ssthresh, 2); // threshold is at least 2 MSS
cwnd = cwnd * beta_cubic; // window reduction cwnd = cwnd * beta_cubic; // window reduction
A side effect of setting beta_cubic to a bigger value than 0.5 is A side effect of setting beta_cubic to a bigger value than 0.5 is
slower convergence. We believe that while a more adaptive setting of slower convergence. We believe that while a more adaptive setting of
beta_cubic could result in faster convergence, it will make the beta_cubic could result in faster convergence, it will make the
analysis of the protocol much harder. This adaptive adjustment of analysis of CUBIC much harder. This adaptive adjustment of
beta_cubic is an item for the next version of CUBIC. beta_cubic is an item for the next version of CUBIC.
4.6. Fast convergence 4.6. Fast convergence
To improve the convergence speed of CUBIC, we add a heuristic in the To improve the convergence speed of CUBIC, we add a heuristic in
protocol. When a new flow joins the network, existing flows in the CUBIC. When a new flow joins the network, existing flows in the
network need to give up their bandwidth shares to allow the flow some network need to give up some of their bandwidth to allow the new flow
room for growth if the existing flows have been using all the some room for growth if the existing flows have been using all the
bandwidth of the network. To increase this release of bandwidth by bandwidth of the network. To speed up this bandwidth release by
existing flows, the following mechanism called fast convergence existing flows, the following mechanism called fast convergence
SHOULD be implemented. SHOULD be implemented.
With fast convergence, when a loss event occurs, before a window With fast convergence, when a congestion event occurs, before the
reduction of congestion window, a flow remembers the last value of window reduction of the congestion window, a flow remembers the last
W_max before it updates W_max for the current loss event. Let us value of W_max before it updates W_max for the current congestion
call the last value of W_max to be W_last_max. event. Let us call the last value of W_max to be W_last_max.
if (W_max < W_last_max){ // should we make room for others if (W_max < W_last_max){ // should we make room for others
W_last_max = W_max; // remember the last W_max W_last_max = W_max; // remember the last W_max
W_max = W_max*(1.0+beta_cubic)/2.0; // further reduce W_max W_max = W_max*(1.0+beta_cubic)/2.0; // further reduce W_max
} else { } else {
W_last_max = W_max // remember the last W_max W_last_max = W_max // remember the last W_max
} }
At a loss event, if the current value of W_max is less than At a congestion event, if the current value of W_max is less than
W_last_max, this indicates that the saturation point experienced by W_last_max, this indicates that the saturation point experienced by
this flow is getting reduced because of the change in available this flow is getting reduced because of the change in available
bandwidth. Then we allow this flow to release more bandwidth by bandwidth. Then we allow this flow to release more bandwidth by
reducing W_max further. This action effectively lengthens the time reducing W_max further. This action effectively lengthens the time
for this flow to increase its window because the reduced W_max forces for this flow to increase its congestion window because the reduced
the flow to have the plateau earlier. This allows more time for the W_max forces the flow to have the plateau earlier. This allows more
new flow to catch up its window size time for the new flow to catch up its congestion window size
The fast convergence is designed for network environments with The fast convergence is designed for network environments with
multiple CUBIC flows. In network environments with only a single multiple CUBIC flows. In network environments with only a single
CUBIC flow and without any other traffic, the fast convergence SHOULD CUBIC flow and without any other traffic, the fast convergence SHOULD
be disabled. be disabled.
4.7. Timeout 4.7. Timeout
In case of timeout, CUBIC follows the standard TCP to reduce cwnd, In case of timeout, CUBIC follows Standard TCP to reduce cwnd
but sets ssthresh using beta_cubic (same as in Section 4.5). [RFC5681], but sets ssthresh using beta_cubic (same as in
Section 4.5) that is different from Standard TCP [RFC5681].
During the first congestion avoidance after a timeout, CUBIC
increases its congestion window size using Eq. 1, where t is the
elapsed time since the beginning of the current congestion avoidance,
K is set to 0, and W_max is set to the congestion window size at the
beginning of the current congestion avoidance.
4.8. Slowstart 4.8. Slowstart
CUBIC MUST employ a slow start algorithm, when the cwnd is no more CUBIC MUST employ a slow start algorithm, when the cwnd is no more
than ssthresh. Among the slow start algorithms, CUBIC MAY choose the than ssthresh. Among the slow start algorithms, CUBIC MAY choose the
standard TCP slow start [RFC5681] in general networks, or the limited standard TCP slow start [RFC5681] in general networks, or the limited
slow start [RFC3742] or hybrid slow start [HR08] for high-bandwidth slow start [RFC3742] or hybrid slow start [HR08] for fast and long-
and long-distance networks. distance networks.
In the case when CUBIC runs the hybrid slow start [HR08], it may exit In the case when CUBIC runs the hybrid slow start [HR08], it may exit
the first slow start without incurring any packet loss and thus W_max the first slow start without incurring any packet loss and thus W_max
is undefined. In this special case, CUBIC switches to congestion is undefined. In this special case, CUBIC switches to congestion
avoidance and increases its congestion window size using Eq. 1 where avoidance and increases its congestion window size using Eq. 1, where
K is set to 0 and W_max is set to the window size when CUBIC just t is the elapsed time since the beginning of the current congestion
exits the slow start. avoidance, K is set to 0, and W_max is set to the congestion window
size at the beginning of the current congestion avoidance.
5. Discussion 5. Discussion
In this section, we further discuss the safety features of CUBIC In this section, we further discuss the safety features of CUBIC
following the guidelines specified in [RFC5033]. following the guidelines specified in [RFC5033].
With a deterministic loss model where the number of packets between With a deterministic loss model where the number of packets between
two successive lost events is always 1/p, CUBIC always operates with two successive packet losses is always 1/p, CUBIC always operates
the concave window profile which greatly simplifies the performance with the concave window profile which greatly simplifies the
analysis of CUBIC. The average window size of CUBIC can be obtained performance analysis of CUBIC. The average window size of CUBIC can
by the following function: be obtained by the following function:
AVG_W_cubic = [C*(3+beta_cubic)/(4*(1-beta_cubic))]^0.25 * AVG_W_cubic = [C*(3+beta_cubic)/(4*(1-beta_cubic))]^0.25 *
(RTT^0.75) / (p^0.75) (Eq. 5) (RTT^0.75) / (p^0.75) (Eq. 5)
With beta_cubic set to 0.7, the above formula is reduced to: With beta_cubic set to 0.7, the above formula is reduced to:
AVG_W_cubic = (C*3.7/1.2)^0.25 * (RTT^0.75) / (p^0.75) (Eq. 6) AVG_W_cubic = (C*3.7/1.2)^0.25 * (RTT^0.75) / (p^0.75) (Eq. 6)
We will determine the value of C in the following subsection using We will determine the value of C in the following subsection using
Eq. 6. Eq. 6.
5.1. Fairness to standard TCP 5.1. Fairness to standard TCP
In environments where standard TCP is able to make reasonable use of In environments where Standard TCP is able to make reasonable use of
the available bandwidth, CUBIC does not significantly change this the available bandwidth, CUBIC does not significantly change this
state. state.
Standard TCP performs well in the following two types of networks: Standard TCP performs well in the following two types of networks:
1. networks with a small bandwidth-delay product (BDP) 1. networks with a small bandwidth-delay product (BDP)
2. networks with a short RTT, but not necessarily a small BDP 2. networks with a short RTT, but not necessarily a small BDP
CUBIC is designed to behave very similarly to standard TCP in the CUBIC is designed to behave very similarly to Standard TCP in the
above two types of networks. The following two tables show the above two types of networks. The following two tables show the
average window size of standard TCP, HSTCP, and CUBIC. The average average window sizes of Standard TCP, HSTCP, and CUBIC. The average
window size of standard TCP and HSTCP is from [RFC3649]. The average window sizes of Standard TCP and HSTCP are from [RFC3649]. The
window size of CUBIC is calculated by using Eq. 6 and CUBIC TCP average window size of CUBIC is calculated using Eq. 6 and the CUBIC
friendly mode for three different values of C. TCP friendly region for three different values of C.
+----------+-------+--------+-------------+-------------+-----------+ +--------+----------+-----------+------------+-----------+----------+
| Loss | TCP | HSTCP | CUBIC | CUBIC | CUBIC | | Loss | Average | Average | CUBIC | CUBIC | CUBIC |
| Rate P | | | (C=0.04) | (C=0.4) | (C=4) | | Rate P | TCP W | HSTCP W | (C=0.04) | (C=0.4) | (C=4) |
+----------+-------+--------+-------------+-------------+-----------+ +--------+----------+-----------+------------+-----------+----------+
| 10^-2 | 12 | 12 | 12 | 12 | 12 | | 10^-2 | 12 | 12 | 12 | 12 | 12 |
| 10^-3 | 38 | 38 | 38 | 38 | 59 | | 10^-3 | 38 | 38 | 38 | 38 | 59 |
| 10^-4 | 120 | 263 | 120 | 187 | 333 | | 10^-4 | 120 | 263 | 120 | 187 | 333 |
| 10^-5 | 379 | 1795 | 593 | 1054 | 1874 | | 10^-5 | 379 | 1795 | 593 | 1054 | 1874 |
| 10^-6 | 1200 | 12279 | 3332 | 5926 | 10538 | | 10^-6 | 1200 | 12279 | 3332 | 5926 | 10538 |
| 10^-7 | 3795 | 83981 | 18740 | 33325 | 59261 | | 10^-7 | 3795 | 83981 | 18740 | 33325 | 59261 |
| 10^-8 | 12000 | 574356 | 105383 | 187400 | 333250 | | 10^-8 | 12000 | 574356 | 105383 | 187400 | 333250 |
+----------+-------+--------+-------------+-------------+-----------+ +--------+----------+-----------+------------+-----------+----------+
Response function of standard TCP, HSTCP, and CUBIC in networks with Response function of Standard TCP, HSTCP, and CUBIC in networks with
RTT = 0.1 seconds. The average window size is in MSS-sized segments. RTT = 0.1 seconds. The average window size is in MSS-sized segments.
Table 1 Table 1
+--------+-----------+-----------+------------+-----------+---------+ +--------+-----------+-----------+------------+-----------+---------+
| Loss | Average | Average | CUBIC | CUBIC | CUBIC | | Loss | Average | Average | CUBIC | CUBIC | CUBIC |
| Rate P | TCP W | HSTCP W | (C=0.04) | (C=0.4) | (C=4) | | Rate P | TCP W | HSTCP W | (C=0.04) | (C=0.4) | (C=4) |
+--------+-----------+-----------+------------+-----------+---------+ +--------+-----------+-----------+------------+-----------+---------+
| 10^-2 | 12 | 12 | 12 | 12 | 12 | | 10^-2 | 12 | 12 | 12 | 12 | 12 |
| 10^-3 | 38 | 38 | 38 | 38 | 38 | | 10^-3 | 38 | 38 | 38 | 38 | 38 |
| 10^-4 | 120 | 263 | 120 | 120 | 120 | | 10^-4 | 120 | 263 | 120 | 120 | 120 |
| 10^-5 | 379 | 1795 | 379 | 379 | 379 | | 10^-5 | 379 | 1795 | 379 | 379 | 379 |
| 10^-6 | 1200 | 12279 | 1200 | 1200 | 1874 | | 10^-6 | 1200 | 12279 | 1200 | 1200 | 1874 |
| 10^-7 | 3795 | 83981 | 3795 | 5926 | 10538 | | 10^-7 | 3795 | 83981 | 3795 | 5926 | 10538 |
| 10^-8 | 12000 | 574356 | 18740 | 33325 | 59261 | | 10^-8 | 12000 | 574356 | 18740 | 33325 | 59261 |
+--------+-----------+-----------+------------+-----------+---------+ +--------+-----------+-----------+------------+-----------+---------+
Response function of standard TCP, HSTCP, and CUBIC in networks with Response function of Standard TCP, HSTCP, and CUBIC in networks with
RTT = 0.01 seconds. The average window size is in MSS-sized RTT = 0.01 seconds. The average window size is in MSS-sized
segments. segments.
Table 2 Table 2
Both tables show that CUBIC with any of these three C values is more Both tables show that CUBIC with any of these three C values is more
friendly to TCP than HSTCP, especially in networks with a short RTT friendly to TCP than HSTCP, especially in networks with a short RTT
where TCP performs reasonably well. For example, in a network with where TCP performs reasonably well. For example, in a network with
RTT = 0.01 seconds and p=10^-6, TCP has an average window of 1200 RTT = 0.01 seconds and p=10^-6, TCP has an average window of 1200
packets. If the packet size is 1500 bytes, then TCP can achieve an packets. If the packet size is 1500 bytes, then TCP can achieve an
average rate of 1.44 Gbps. In this case, CUBIC with C=0.04 or C=0.4 average rate of 1.44 Gbps. In this case, CUBIC with C=0.04 or C=0.4
achieves exactly the same rate as Standard TCP, whereas HSTCP is achieves exactly the same rate as Standard TCP, whereas HSTCP is
about ten times more aggressive than Standard TCP. about ten times more aggressive than Standard TCP.
We can see that C determines the aggressiveness of CUBIC in competing We can see that C determines the aggressiveness of CUBIC in competing
with other protocols for the bandwidth. CUBIC is more friendly to with other congestion control algorithms for the bandwidth. CUBIC is
the Standard TCP, if the value of C is lower. However, we do not more friendly to the Standard TCP, if the value of C is lower.
recommend to set C to a very low value like 0.04, since CUBIC with a However, we do not recommend to set C to a very low value like 0.04,
low C cannot efficiently use the bandwidth in long RTT and high since CUBIC with a low C cannot efficiently use the bandwidth in long
bandwidth networks. Based on these observations and our experiments, RTT and high bandwidth networks. Based on these observations and our
we find C=0.4 gives a good balance between TCP-friendliness and experiments, we find C=0.4 gives a good balance between TCP-
aggressiveness of window growth. Therefore, C SHOULD be set to 0.4. friendliness and aggressiveness of window increase. Therefore, C
With C set to 0.4, Eq. 6 is reduced to: SHOULD be set to 0.4. With C set to 0.4, Eq. 6 is reduced to:
AVG_W_cubic = 1.054 * (RTT^0.75) / (p^0.75) (Eq. 7) AVG_W_cubic = 1.054 * (RTT^0.75) / (p^0.75) (Eq. 7)
Eq. 7 is then used in the next subsection to show the scalability of Eq. 7 is then used in the next subsection to show the scalability of
CUBIC. CUBIC.
5.2. Using Spare Capacity 5.2. Using Spare Capacity
CUBIC uses a more aggressive window growth function than Standard TCP CUBIC uses a more aggressive window increase function than Standard
under long RTT and high bandwidth networks. TCP under long RTT and high bandwidth networks.
The following table shows that to achieve 10Gbps rate, standard TCP The following table shows that to achieve the 10Gbps rate, Standard
requires a packet loss rate of 2.0e-10, while CUBIC requires a packet TCP requires a packet loss rate of 2.0e-10, while CUBIC requires a
loss rate of 2.9e-8. packet loss rate of 2.9e-8.
+------------------+-----------+---------+---------+---------+ +------------------+-----------+---------+---------+---------+
| Throughput(Mbps) | Average W | TCP P | HSTCP P | CUBIC P | | Throughput(Mbps) | Average W | TCP P | HSTCP P | CUBIC P |
+------------------+-----------+---------+---------+---------+ +------------------+-----------+---------+---------+---------+
| 1 | 8.3 | 2.0e-2 | 2.0e-2 | 2.0e-2 | | 1 | 8.3 | 2.0e-2 | 2.0e-2 | 2.0e-2 |
| 10 | 83.3 | 2.0e-4 | 3.9e-4 | 2.9e-4 | | 10 | 83.3 | 2.0e-4 | 3.9e-4 | 2.9e-4 |
| 100 | 833.3 | 2.0e-6 | 2.5e-5 | 1.4e-5 | | 100 | 833.3 | 2.0e-6 | 2.5e-5 | 1.4e-5 |
| 1000 | 8333.3 | 2.0e-8 | 1.5e-6 | 6.3e-7 | | 1000 | 8333.3 | 2.0e-8 | 1.5e-6 | 6.3e-7 |
| 10000 | 83333.3 | 2.0e-10 | 1.0e-7 | 2.9e-8 | | 10000 | 83333.3 | 2.0e-10 | 1.0e-7 | 2.9e-8 |
+------------------+-----------+---------+---------+---------+ +------------------+-----------+---------+---------+---------+
skipping to change at page 12, line 49 skipping to change at page 13, line 9
Table 3 Table 3
Our test results in [HKLRX06] indicate that CUBIC uses the spare Our test results in [HKLRX06] indicate that CUBIC uses the spare
bandwidth left unused by existing Standard TCP flows in the same bandwidth left unused by existing Standard TCP flows in the same
bottleneck link without taking away much bandwidth from the existing bottleneck link without taking away much bandwidth from the existing
flows. flows.
5.3. Difficult Environments 5.3. Difficult Environments
CUBIC is designed to remedy the poor performance of TCP in fast long- CUBIC is designed to remedy the poor performance of TCP in fast and
distance networks. long-distance networks.
5.4. Investigating a Range of Environments 5.4. Investigating a Range of Environments
CUBIC has been extensively studied by using both NS-2 simulation and CUBIC has been extensively studied by using both NS-2 simulation and
test-bed experiments covering a wide range of network environments. test-bed experiments covering a wide range of network environments.
More information can be found in [HKLRX06]. More information can be found in [HKLRX06].
Same as Standard TCP, CUBIC is a loss-based congestion control Same as Standard TCP, CUBIC is a loss-based congestion control
algorithm. Because CUBIC is designed to be more aggressive (due to algorithm. Because CUBIC is designed to be more aggressive (due to
faster window growth function and bigger multiplicative decrease faster window increase function and bigger multiplicative decrease
factor) than Standard TCP in fast and long distance networks, it can factor) than Standard TCP in fast and long distance networks, it can
fill large drop-tail buffers more quickly than Standard TCP and fill large drop-tail buffers more quickly than Standard TCP and
increase the risk of a standing queue[KWAF16]. In this case, proper increase the risk of a standing queue[KWAF16]. In this case, proper
queue sizing and management [RFC7567] could be used to reduce the queue sizing and management [RFC7567] could be used to reduce the
packet queueing delay. packet queueing delay.
5.5. Protection against Congestion Collapse 5.5. Protection against Congestion Collapse
With regard to the potential of causing congestion collapse, CUBIC With regard to the potential of causing congestion collapse, CUBIC
behaves like standard TCP since CUBIC modifies only the window behaves like Standard TCP since CUBIC modifies only the window
adjustment algorithm of TCP. Thus, it does not modify the ACK adjustment algorithm of TCP. Thus, it does not modify the ACK
clocking and Timeout behaviors of Standard TCP. clocking and Timeout behaviors of Standard TCP.
5.6. Fairness within the Alternative Congestion Control Algorithm. 5.6. Fairness within the Alternative Congestion Control Algorithm.
CUBIC ensures convergence of competing CUBIC flows with the same RTT CUBIC ensures convergence of competing CUBIC flows with the same RTT
in the same bottleneck links to an equal bandwidth share. When in the same bottleneck links to an equal throughput. When competing
competing flows have different RTTs, their bandwidth shares are flows have different RTTs, their throughput ratio is linearly
linearly proportional to the inverse of their RTT ratios. This is proportional to the inverse of their RTT ratios. This is true
true independent of the level of statistical multiplexing in the independent of the level of statistical multiplexing in the link.
link.
5.7. Performance with Misbehaving Nodes and Outside Attackers 5.7. Performance with Misbehaving Nodes and Outside Attackers
This is not considered in the current CUBIC. This is not considered in the current CUBIC.
5.8. Behavior for Application-Limited Flows 5.8. Behavior for Application-Limited Flows
CUBIC does not raise its congestion window size if the flow is CUBIC does not raise its congestion window size if the flow is
currently limited by the application instead of the congestion currently limited by the application instead of the congestion
window. In case of long periods when cwnd has not been updated due window. In case of long periods when cwnd has not been updated due
skipping to change at page 14, line 12 skipping to change at page 14, line 14
NOT include these periods; otherwise, W_cubic(t) might be very high NOT include these periods; otherwise, W_cubic(t) might be very high
after restarting from these periods. after restarting from these periods.
5.9. Responses to Sudden or Transient Events 5.9. Responses to Sudden or Transient Events
In case that there is a sudden congestion, a routing change, or a In case that there is a sudden congestion, a routing change, or a
mobility event, CUBIC behaves the same as Standard TCP. mobility event, CUBIC behaves the same as Standard TCP.
5.10. Incremental Deployment 5.10. Incremental Deployment
CUBIC requires only the change of TCP senders, and does not require CUBIC requires only the change of TCP senders, and it does not make
any assistant of routers. any changes to TCP receivers. That is, a CUBIC sender works
correctly with the Standard TCP receivers. In addition, CUBIC does
not require any changes to the routers, and does not require any
assistant from the routers.
6. Security Considerations 6. Security Considerations
This proposal makes no changes to the underlying security of TCP. This proposal makes no changes to the underlying security of TCP.
More information about TCP security concerns can be found in
[RFC5681].
7. IANA Considerations 7. IANA Considerations
There are no IANA considerations regarding this document. There are no IANA considerations regarding this document.
8. Acknowledgements 8. Acknowledgements
Alexander Zimmermann and Lars Eggert have received funding from the Alexander Zimmermann and Lars Eggert have received funding from the
European Union's Horizon 2020 research and innovation program European Union's Horizon 2020 research and innovation program
2014-2018 under grant agreement No. 644866 (SSICLOPS). This document 2014-2018 under grant agreement No. 644866 (SSICLOPS). This document
skipping to change at page 14, line 41 skipping to change at page 14, line 48
9. References 9. References
9.1. Normative References 9.1. 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, Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997, DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>. <https://www.rfc-editor.org/info/rfc2119>.
[RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition
of Explicit Congestion Notification (ECN) to IP",
RFC 3168, DOI 10.17487/RFC3168, September 2001,
<https://www.rfc-editor.org/info/rfc3168>.
[RFC3649] Floyd, S., "HighSpeed TCP for Large Congestion Windows", [RFC3649] Floyd, S., "HighSpeed TCP for Large Congestion Windows",
RFC 3649, DOI 10.17487/RFC3649, December 2003, RFC 3649, DOI 10.17487/RFC3649, December 2003,
<https://www.rfc-editor.org/info/rfc3649>. <https://www.rfc-editor.org/info/rfc3649>.
[RFC3742] Floyd, S., "Limited Slow-Start for TCP with Large [RFC3742] Floyd, S., "Limited Slow-Start for TCP with Large
Congestion Windows", RFC 3742, DOI 10.17487/RFC3742, March Congestion Windows", RFC 3742, DOI 10.17487/RFC3742, March
2004, <https://www.rfc-editor.org/info/rfc3742>. 2004, <https://www.rfc-editor.org/info/rfc3742>.
[RFC4960] Stewart, R., Ed., "Stream Control Transmission Protocol", [RFC4960] Stewart, R., Ed., "Stream Control Transmission Protocol",
RFC 4960, DOI 10.17487/RFC4960, September 2007, RFC 4960, DOI 10.17487/RFC4960, September 2007,
skipping to change at page 16, line 22 skipping to change at page 16, line 35
[K03] Kelly, T., "Scalable TCP: Improving Performance in [K03] Kelly, T., "Scalable TCP: Improving Performance in
HighSpeed Wide Area Networks", ACM SIGCOMM Computer HighSpeed Wide Area Networks", ACM SIGCOMM Computer
Communication Review , April 2003. Communication Review , April 2003.
[KWAF16] Khademi, N., Welzl, M., Armitage, G., and G. Fairhurst, [KWAF16] Khademi, N., Welzl, M., Armitage, G., and G. Fairhurst,
"TCP Alternative Backoff with ECN (ABE)", Internet-draft, "TCP Alternative Backoff with ECN (ABE)", Internet-draft,
IETF work-in-progress draft-khademi-tcpm- IETF work-in-progress draft-khademi-tcpm-
alternativebackoff-ecn-01 , October 2016. alternativebackoff-ecn-01 , October 2016.
[LS08] Leith, D. and R. Shorten, "H-TCP: TCP Congestion Control
for High Bandwidth-Delay Product Paths", Internet-draft
draft-leith-tcp-htcp-06 , April 2008.
[XHR04] Xu, L., Harfoush, K., and I. Rhee, "Binary Increase [XHR04] Xu, L., Harfoush, K., and I. Rhee, "Binary Increase
Congestion Control for Fast, Long Distance Networks", In Congestion Control for Fast, Long Distance Networks", In
Proceedings of IEEE INFOCOM , March 2004. Proceedings of IEEE INFOCOM , March 2004.
Authors' Addresses Authors' Addresses
Injong Rhee Injong Rhee
North Carolina State University North Carolina State University
Department of Computer Science Department of Computer Science
Raleigh, NC 27695-7534 Raleigh, NC 27695-7534
 End of changes. 68 change blocks. 
255 lines changed or deleted 284 lines changed or added

This html diff was produced by rfcdiff 1.46. The latest version is available from http://tools.ietf.org/tools/rfcdiff/