a simulation based comparison between highspeed tcp and xcp
DESCRIPTION
A Simulation Based Comparison Between HighSpeed TCP and XCP. Jae Wook Lee and Gleb Chuvpilo 6.829 Final Project December 6, 2002. Talk at a Glance. Motivation Overview of HSTCP and XCP Our Results Utilization, Fairness, TCP-friendliness, Dynamics, Buffering, Deployment Conclusion - PowerPoint PPT PresentationTRANSCRIPT
A Simulation Based Comparison Between HighSpeed TCP and XCP
Jae Wook Lee and Gleb Chuvpilo
6.829 Final Project
December 6, 2002
2
Talk at a Glance
1. Motivation2. Overview of HSTCP and XCP3. Our Results
Utilization, Fairness, TCP-friendliness, Dynamics, Buffering, Deployment
4. Conclusion5. Future Work
3
Motivation For a TCP throughput of 10 Gbps:
Need cwnd=80,000 packets Increase cwnd by 1 packet/RTT
Takes thousands of RTT to ramp up to full utilization!
HighSpeed TCP and XCP are two proposed replacements
Which one is better and when? What are the tradeoffs?
4
HighSpeed TCP (HSTCP) Modify TCP response function when cwnd is
high to:
The whole point is that a(w) increases and b(w) decreases as cwnd becomes larger.
Example: behavior when cwnd = 80,000 packets:
5.0
loglog
loglog5.0
rate) droppacket :)(( 2
2)(2
LowHigh
Low
WW
WwseHighDecreawb
wpwb
wbwpwwa
TCP HSTCP
a(w) 1 72
b(w) 0.5 0.1
5
eXplicit Control Protocol (XCP)
Routers provide explicit feedback to senders No per-flow state is maintained in routers Decoupled fairness and efficiency controllers
Efficiency controller increases aggregate feedback proportionally to spare bandwidth and decreases proportionally to the persistent queue size
Fairness controller uses AIMD
6
Simulation Topology
SF0
SF1
SFn-1
R0 R1
SR0
SR1
SRm-1
Dumbbell TopologynFlows = variableRTT=100msBW=100Mbps (variable)qType=RED/DropTailqSize=BWDelay
7
Utilization Do HSTCP and XCP achieve high utilization in
high bandwidth-delay networks?
Bottleneck util. vs. variable BW Bottleneck util. vs. variable RTTSETUP RTT=(100ms)
Bottleneck BW=(200Mbps)Qsize=BW*RTT
8
Fairness (Jain’s Index) A fairness index proposed by Jain:
ii
ii
xn
x
xF2
2
)(
• For n flows:- Best Fairness F(x) = 1 - Worst Fairness F(x) = 1/n
Fairness Index vs. Number of FlowsSETUP RTT=100ms
Bottleneck BW=200MbpsQsize=BW*RTT
9
TCP-friendliness HighSpeed TCP flow starves a TCP flow even in
relatively low/moderate bandwidth (e. g. 50Mbps)
Starvation of TCP flow(>10x)
SETUP RTT=100msBottleneck BW=50Mbps
Qsize=BW*RTTQtype=DropTail
Standard TCP flows fair to each
other
1 HSTCP and 1 TCP flow 2 TCP flows
10
Dynamics (Convergence to Fairness)
XCPHSTCP
1. In general, HSTCP takes longer time to converge to fairness than XCP2. HSTCP may take longer time to converge than TCP, too.
Long convergence time to fairness (~4000 RTTs)
Very short convergence time to fairness
SETUP RTT=100msBottleneck BW=50Mbps
Qsize=BW*RTTQtype=DropTail
A new flow joins at t=40s
TCP
Long convergence time to fairness (~1700 RTTs)
11
Dynamics (Fetching BW)
XCPHSTCP
Takes ~310 RTTs to grab suddenly available bandwidth
HSTCP may rely on additive increase to fetch suddenly available bandwidth like TCP
Immediately grabs suddenly available bandwidth
SETUP RTT=100msBottleneck BW=50Mbps
Qsize=BW*RTTQtype=DropTail
TCP
Takes ~2550 RTTs to grab suddenly available bandwidth
12
Buffering (Utilization) What is the impact of buffering on
utilization?
SETUP RTT=100msBottleneck BW=200Mbps
NumFlows=10
*Buffer size is normalized to bandwidth-delay product
13
Buffering (Fairness) What is the impact of buffering on fairness?
qSize = 0.1 BW Delay
XCPHSTCP SETUP RTT=100msBottleneck BW=200Mbps
Qsize=0.1*BW*RTTQtype=DropTail/RED
14
Effects of Reverse Flows Two new phenomena present [Zhang, et al]
ACK-compression Out-of-phase queue-synchronization
Reverse flows degrades min-max fairness (by factor of ~2) of HSTCP flows as well as utilization.
15
Deployment What are the costs and side effects of
deployment? HSTCP
Easier to deploy – end-to-end protocol, no router support needed.
HighSpeed TCP flows starve TCP flows XCP:
Harder to deploy – requires router support on the path.
Next generation of IP routers may accommodate XCP with a benefit of smaller buffering.
16
Conclusion Both HSTCP and XCP demonstrate good utilization for high
bandwidth-delay environments. Fairness index: XCP > HSTCP > TCP, but all close to 1. HSTCP starves TCP flows even in low-bandwidth networks. Dynamics tradeoff: convergence to fairness vs.
convergence to full utilization. With reasonably small buffer size of 0.1×BW×Delay, both
HSTCP and XCP work well in terms of utilization and fairness. But, XCP outperforms HSTCP for both criteria.
Both HSTCP and XCP face deployment problems: HSTCP is TCP-unfriendly, XCP needs router support on the path.
17
Questions?
18
Future Work
Evaluate QuickStart Obtain results for TCP friendliness of
XCP Simulate Web traffic Explain specific convergence issues with
HSTCP Analyze and in XCP
Appendix
19
Fairness (Dynamics) Are protocols scalable in terms of
fairness in a homogeneous environment of the future?
XCPHSTCP
Appendix
20
Dynamics (HSTCP and TCP)
Jain’s analysis for HSTCP
“Convergence to fairness” condition holds in HighSpeed TCP
Theta() analysis
For =the angle between a(w) & wb(w)
small quickly fetch available bw but, could cause slow conv. to
fairness
There is a tradeoff between convergence to fairness and efficiency in achieving full utilization.
1)(2
)(2 ),(),(
1
2
2.0
1
22121
wb
wb
w
wwwww DI
Efficiency line
Fairnessline
cwnd1
cwnd2
(w1,w2)
DI
HSTCP
TCP
Appendix
21
Theta analysis (Details)
Appendix
In order to converge to fairness fast:
Efficiency line
Fairnessline
cwnd1
cwnd2
(w1,w2)
DI
1. The larger D, the better
2. The smaller I, the better (I>45 always)
3. Combining 1 & 2, maximize tanD / tanI.
11
22tanwbw
wbwD
1))(2()(
))(2()(
))(2()(2078.0
))(2()(2078.0
)(
)(tan
218.0
1
128.0
2
212.1
221
122.1
122
1
2
wbwbw
wbwbw
wbwbww
wbwbww
wa
waI
1)(2
)(2
tan
tan
1
2
2.0
1
2
wb
wb
w
w
I
D