1 Long-Run Behavior of Equation-Based Rate Control: Theory and its Empirical Validation Milan Vojnović Seminar on Theory of Communication Networks, ETHZ, Zürich, May 6, 2003 Joint work with Jean-Yves Le Boudec Lab and Internet measurements with C. Laetsch, T. Müller
40
Embed
Long-Run Behavior of Equation-Based Rate Control: Theory and its Empirical Validation
Long-Run Behavior of Equation-Based Rate Control: Theory and its Empirical Validation. Milan Vojnović. Joint work with Jean-Yves Le Boudec Lab and Internet measurements with C. Laetsch, T. Müller. Seminar on Theory of Communication Networks, ETHZ, Zürich, May 6, 2003. My thesis. - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Long-Run Behavior of Equation-Based Rate
Control:Theory and its Empirical Validation
Milan Vojnović
Seminar on Theory of Communication Networks, ETHZ, Zürich, May 6, 2003
Joint work with Jean-Yves Le BoudecLab and Internet measurements with C. Laetsch, T.
Müller
2
My thesis
equation-based rate control -- is it TCP friendly ?
increase-decrease controls -- e.g. TCP-- fairness in bandwidth sharing
expedited forwarding-- queueing bounds for diffserv EF
input-queued switch-- scheduler latency
This talk:
3
Problem we study TCP -- Internet predominant transport protocol;
implements a window-based transmission control Equation-based rate control -- rate-based
transmission control (e.g. for media streaming)-- TFRC (TCP-Friendly Rate Control)
Floyd et al (2000), an IETF internet-draft Controls need to be TCP-friendly -- an axiom
established by part of Internet research community (mid-nineties)
TCP TCP
Internetnon-TCP
non-TCP
4
Problem we study (cont’d)TCP characterized by:
TCP throughput = f(loss-event rate)Basic control law of equation-based
rate control: loss-event rate estimated on-line
(call the estimator ) at some instants
send rate = Where is the problem ?
f is non-linear, loss is random
sampling bias-- rate set at special points in time )p̂(f
• TFRC = experimental code (ICIR, 2000), we adapted to conform to TFRC spec
• Background = equal # of TCPs and TFRCs
• R = UMASS, INRIA, Melbourne, Caltech, KTH, Hong Kong
Setting:
100 Mb/s
10 or 100 Mb/s
Internet
R
Slides: R = UMASSAccess at R = 100 Mb/s
32
Internet measurements: EPFL -> UMASS
(C) Is the control conservative ?
)r̂,p̂(f
x̂
TFRCTFRC
TFRC
TFRCp̂
=> yes
(F) Does TCP conform to its formula ?
=> not always
)r̂,p̂(f
x̂
TCPTCP
TCP
TCPp̂
(P) Is loss-rate no better than TCP’s ?
experiment
=> not alwaysTCP
TFRC
p̂
p̂
33(pkts/s) x̂TCP
(pkts/s) x̂TFRC => no
(TCP-f) Is the control TCP-friendly ?
Internet measurements: EPFL -> UMASS
both, (P) and (F) not true
34
Lab experiments
TCP TFRC
10 Mb/s
Background
qdisc = RED, Droptail
delay= 50 ms
Circles = PCs, Linux kernel 2.4.18
Setting:
• TCP, TFRC, Background = same as with lab experiments
• Delay = emulated by NIST Net100 Mb/s
100 Mb/s
35
Lab experiments with RED (cont’d)(C) Is the control conservative ?
)r̂,p̂(f
x̂
TFRCTFRC
TFRC
TFRCp̂
=> yes
(F) Does TCP conform to its formula ?
=> no, mostly overshoots
)r̂,p̂(f
x̂
TCPTCP
TCP
TCPp̂
(P) Is loss-rate no better than TCP’s ?
experiment
=> not alwaysTCP
TFRC
p̂
p̂
36
=> yes
Lab experiments with RED (cont’d)
(pkts/s) x̂TCP
(pkts/s) x̂TFRC
(TCP-f) Is the control TCP-friendly ?
37
Lab experiments with DropTail (100 pkts)
(C) Is the control conservative ?
)r̂,p̂(f
x̂
TFRCTFRC
TFRC
TFRCp̂
=> yes
(F) Does TCP conform to its formula ?
=> no)r̂,p̂(f
x̂
TCPTCP
TCP
TCPp̂
(P) Is loss-rate no better than TCP’s ?
=> yesTCP
TFRC
p̂
p̂
experiment
38
(pkts/s) x̂TCP
(pkts/s) x̂TFRC
=> not alwaysif yes, mostlyexcessively
Lab experiments with DropTail (100 pkts)
(TCP-f) Is the control TCP-friendly ?
(P) true, but large discrepancy
39
Separate factors ! (C) conditions for either conservative
or non-conservative control-- TFRC throughput-drop explained-- a control with PFTK and fixed packet send rate intrinsically non-conservative for large loss-event rate
(P) in many-sources regime, expect loss-event rate be larger than TCP sees-- other regimes exist where (P) is not true
(F) TCP may deviate from PFTK formula
Conclusion
40
variability of round-trip time, its correlation with loss process -- do they matter ?
conservativeness -- seek for realistic cases when the control is non-conservative
loss-event rate-- when and why it is smaller (or larger) than TCP’s ?