Is the Round- -trip Time trip Time Is the Round Correlated with - - PowerPoint PPT Presentation

is the round trip time trip time is the round correlated
SMART_READER_LITE
LIVE PREVIEW

Is the Round- -trip Time trip Time Is the Round Correlated with - - PowerPoint PPT Presentation

Is the Round- -trip Time trip Time Is the Round Correlated with the Number of Correlated with the Number of Packets in Flight? Packets in Flight? Authors : Authors Saad Biaz Nitin H. Vaidya biazsaad@auburn.edu nhv@uiuc.edu Auburn


slide-1
SLIDE 1

Computer Science and Software Engineering, Auburn University 1

Is the Round Is the Round-

  • trip Time

trip Time Correlated with the Number of Correlated with the Number of Packets in Flight? Packets in Flight?

Authors Authors: Saad Biaz biazsaad@auburn.edu Auburn University Nitin H. Vaidya nhv@uiuc.edu University of Illinois at Urbana-Champaign Presenter Presenter: : Xia Wang wangxi2@auburn.edu Auburn University

slide-2
SLIDE 2

Computer Science and Software Engineering, Auburn University 2

Motivation

Some Congestion Avoidance Techniques implicitly

use the correlation between RTT and the number

  • f packets in flight.
  • TCP-Vegas
  • Z. Wang and J. Crowroft “A New Congestion Control Scheme:

Slow Start and Search (Tri-S)”

  • Raj Jain “A Delay-based Approach for Congestion Avoidance in

Interconnected Heterogeneous Computer Networks”

  • S. Biaz and N. H. Vaidya. “Distinguishing Congestion Losses

from Wireless Transmission Losses”

slide-3
SLIDE 3

Computer Science and Software Engineering, Auburn University 3

What can we do

Sample coefficient of correlation Xi: RTT Yi: the number of packets in flight

∑ ∑ ∑

− − − −

i i i i i i i

y y x x y y x x

2 2

) ( ) ( ) )( (

ρ(Cl, xi,yi) =

slide-4
SLIDE 4

Computer Science and Software Engineering, Auburn University 4

Notations and Terms

slide-5
SLIDE 5

Computer Science and Software Engineering, Auburn University 5

Data Collection

Data is based on Dr. Vern Paxon’s PhD thesis

“Measurements and Analysis of End-to-End Internet Dynamics”

Tcpdump traces collected over 37 sites,

14,218 TCP connections span 737 paths

slide-6
SLIDE 6

Computer Science and Software Engineering, Auburn University 6

Slow Path & Fast Path

64 KBytes/s Frequency Distribution of Connections by Bottleneck Link Speed (Dr. Paxon)

slide-7
SLIDE 7

Computer Science and Software Engineering, Auburn University 7

Correlation between RTT and W

(0.8, 0.26) (0.8, 0.11)

slide-8
SLIDE 8

Computer Science and Software Engineering, Auburn University 8

ρwv(Cl,

RTTi RTTi δ δ

Wi Wi δ δ

,

)

∑ ∑ ∑

− − − −

i i i i i i i

y y x x y y x x

2 2

) ( ) ( ) )( (

ρ(Cl, xi,yi) =

slide-9
SLIDE 9

Computer Science and Software Engineering, Auburn University 9

Correlation in Direction of Change

slide-10
SLIDE 10

Computer Science and Software Engineering, Auburn University 10

Coefficient of Correlation on a Path?

Coefficient of correlation for RTT and W for slow path Coefficient of correlation for direction of change for slow path

slide-11
SLIDE 11

Computer Science and Software Engineering, Auburn University 11

Conclusion and Future Work

No strong correlation between the variation of

RTT and the variations of congestion window size.

Most of TCP connections have positive

correlation in the direction of change of RTT and W.

Updating data collections

slide-12
SLIDE 12

Computer Science and Software Engineering, Auburn University 12

Resources

http://www.crhc.uiuc.edu/~nhv/old.papers/mobile- computing/99-006.ps.Z http://www.eng.auburn.edu/users/sbiaz/tech99-006.ps