ENSC 835: HIGH-PERFORMANCE NETWORKS CMPT 885: SPECIAL TOPICS: - - PowerPoint PPT Presentation

▶
ensc 835 high performance networks cmpt 885 special
SMART_READER_LITE
LIVE PREVIEW

ENSC 835: HIGH-PERFORMANCE NETWORKS CMPT 885: SPECIAL TOPICS: - - PowerPoint PPT Presentation

Sep 25, 2023 280 likes •483 views

ENSC 835: HIGH-PERFORMANCE NETWORKS CMPT 885: SPECIAL TOPICS: HIGH-PERFORMANCE NETWORKS FINAL PROJECT PRESENTATIONS Spring 2006 PROJECT TCP Fairness Analysis of CUBIC TCP Simulated by NS-2 Qing Chen E-mail qingc@sfu.ca 1 Road Map

slide-1
SLIDE 1

1

FINAL PROJECT PRESENTATIONS Spring 2006 PROJECT TCP Fairness Analysis of CUBIC TCP Simulated by NS-2 Qing Chen E-mail qingc@sfu.ca

ENSC 835: HIGH-PERFORMANCE NETWORKS CMPT 885: SPECIAL TOPICS: HIGH-PERFORMANCE NETWORKS

slide-2
SLIDE 2

2

Road Map

Introduction Advanced TCP for High speed network

Approaches BIC and CUBIC TCP Brief review of TCP NEW RENO

TCP Fairness

Fairness Effects of Queuing Management

Simulation Conclusions Future Work Reference

slide-3
SLIDE 3

3

Introduction

High Speed Network

The bandwidth of network rises up to 10Gbps The network covers the long distance ESNet, Abilene

Current TCP faces difficulties in high speed

network

Efficiency degrades when bandwidth-delay product

increases

Oscillation problem Problems with short flows

slide-4
SLIDE 4

4

Approaches

Improve congestion control based on current TCP

HSTCP, STCP HSTCP-S. Floyd, AIMD-based, STCP-Tom Kelly, MIMD-based Fast TCP Steven H. Low’s team, California Institute of Technology, delay based BIC TCP, CUBIC TCP Injong Rhee and Lisong Xu, North Carolina State University HTCP

  • D. Leith and R.N. Shorten, Hamilton Institute. Two modes: High speed and slow

speed.

Others

SABUL XCP

slide-5
SLIDE 5

5

BIC TCP

  • Binary search
  • 1. When loss happens, calculate CWNDmin after loss using multiplicative

decrease, CWND=CWNDmin

  • 2. Calculate a mid point between CWNDmax (CWND just before loss) and

CWNDmin

  • 3. Set the mid point as target, Target=(CWNDmas+CWNDmin)/2
  • 4. If Target-CWND>Max setting, CWND=CWND+Max setting
  • 5. If Target-CWND<=Max Setting, CWND=Target, CWNDmin =Target,

repeat step 2 to 5

  • 6. If CWNDmax-CWNDmin<Min Setting, CWND=CWNDmax and Binary

search completes

  • Two Stages
  • Max. Probe stage
  • Binary search stage
  • Window Growth Pattern

(http://www.csc.ncsu.edu/faculty/rhe e/export/bitcp/index.htm)

slide-6
SLIDE 6

6

CUBIC TCP

Cubic TCP is developed on BIC TCP Main reasons

The window control algorithm of BIC is too

complicated.

BIC TCP could be too aggressive in slow network

with short round trip time.

Improvements

Use a cubic function to search CWnd Involve a elapsed time since last loss when calculate

CWnd

slide-7
SLIDE 7

7

CUBIC TCP (Continue)

Algorithm

When receive an ACK When loss happens

Window Growth Pattern

( )

max 3

cwnd K t C cwnd + − ←

3 max

/C cwnd K β =

max

cwnd cwnd × ← β

(http://www.csc.ncsu.edu/faculty/rhe e/export/bitcp/index.htm)

slide-8
SLIDE 8

8

Brief Review of TCP New Reno

TCP New Reno

Traditional congestion control algorithm AIMD Four Stages

Slow Start Congestion Avoidance Fast retransmit Fast recovery

slide-9
SLIDE 9

9

TCP Fairness

TCP Fairness

When n flows connect to one link, each flow will share 1/n of

total bandwidth

Fairness ratio of two flows Factors that affect Fairness

RTT: Throughput ratio of two flows is inversely proportional to

the ratio of their RTTs

Queuing management Link Capability

j i ij

Thru Thru FR / =

slide-10
SLIDE 10

10

Queuing Management’ Effect

  • n TCP Fairness

RED (Random Early Detection)

Algorithm: Queue model Effects on fairness

Flow with larger sending rate will have higher drop possibility

slide-11
SLIDE 11

11

(Continue)

Drop Tail

Algorithm

No packet drop if queue is not full Drop all packet if queue is full

Queue model Effects on fairness

Flow with higher sending rate will have more

packets in queue

slide-12
SLIDE 12

12

Simulation

Tool: NS-2.26 with CYGWIN Topology and assumption:

Two flows have same RRT

What will be analyzed

Congestion window, throughput, fairness and link

utility

Scenarios and cases

  • 1. Validation
  • 2. RED and Drop tail
  • 3. Three cases with different start time.
  • 4. Two different bandwidths: 20MBPS and 1GBPS

S2 S1 N1 R1 N2 R2

slide-13
SLIDE 13

13

Validation

CWND of New Reno 50 100 150 200 250 1 446 891 1336 1781 2226 2671 3116 3561 Time (100ms) CWND (Packets)

CWND of CUBIC 50 100 150 200 250 1 369 737 1105 1473 1841 2209 2577 2945 3313 3681 Time (100ms) CWND (Packets)

Congestion Window Growth w.r.t. time:

slide-14
SLIDE 14

14

(Continue)

Throughput of New Reno

5 10 15 20 25 1 37 73 109 145 181 217 253 289 325 361 397 Thousands Time (s) Throughput (KB) Throughput of CUBIC 5 10 15 20 25 1 36 71 106 141 176 211 246 281 316 351 386 Thousands Time (s) Throughput (KB)

Link Utility: The average link utility of Cubic is 99.4%, the average of link utility of New Reno is 89.4%

slide-15
SLIDE 15

15

Simulation Result

Fairness:

1GBPS bandwidth,

100ms RTT

With RED: 0.09 With DropTail: 0.14

20MB bandwidth,

100ms RTT

With RED: 0.69 With DropTail: 0.40

Fairness Ratio (New Reno/CUBIC)

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 200 250 300 350 400 450 Time (s) Fairness Ratio DropTail RED

Fairness Rate (Cubic to New Reno)

0.5 1 1.5 2 200 250 300 350 400 450 Time (s) Fairness DropTail RED

slide-16
SLIDE 16

16

Conclusion

The performance of CUBIC is better than that of

TCP New Reno in oscillation and link utility

In middle speed network, TCP fairness of

CUBIC is better than that in high speed network

Fairness will be a problem for CUBIC in high

speed network

Fairness of CUBIC with RED is a little bit better

than that with DropTail in middle speed network but a bit worse in high speed network

slide-17
SLIDE 17

17

Future Work

Analyze fairness performance with other

configuration of TCP parameters and queue type

Compare fairness of other advanced TCP

protocol

slide-18
SLIDE 18

18

References

  • [1] H-TCP: A framework for congestion control in high-speed and long-

distance networks, D.J. Leith, R.N. Shorten, Y.Lee, HI Technical Report, August 2005, http://www.hamilton.ie/net/htcp/

  • [2] Binary Increase Congestion Control for Fast, Long Distance Networks,

Lisong Xu, Khaled Harfoush, and Injong Rhee, ”, In Proceedings of the IEEE INFOCOM, March 2004

  • [3] CUBIC: A New TCP-Friendly High-Speed TCP Variant, Injong Rhee, and

Lisong Xu, PFLDnet 2005, February 2005

  • [4] H-TCP: TCP for high-speed and long-distance networks, D. Leith, R.

Shorten, Second International Workshop on Protocols for Fast Long- Distance Networks, February 16-17, 2004, Argonne, Illinois USA

  • [5] A Study of TCP Fairness in High-Speed Networks, Junsoo Lee, Stephan

Bohacek, Joao P. Hespanha, Katia Obraczka, submitted to ICNP, 2005

  • [6] Congestion Control for High Bandwidth-Delay Product Network, Dina

Katabi, Mark Handley, Charlie Rohrs, ACM SIGCOMM Computer Communication Review, Volume 32, Issue 4, October 2002

slide-19
SLIDE 19

19

Reference (Continue)

  • [7] Fast TCP Motivation, Architecture, Algorithms, Performance, Chen Jin,

David X. Wei, Steven H. Law, In Proceedings of IEEE INFOCOM, March 2004 http://netlab.caltech.edu

  • [8] BIC TCP, http://www.csc.ncsu.edu/faculty/rhee/export/bitcp/
  • [9] The Network Simulator - ns-2, http://www.isi.edu/nsnam/ns/