SELF-SIMILAR INTERNET SELF-SIMILAR INTERNET TRAFFIC AND IMPLICATIONS TRAFFIC AND IMPLICATIONS FOR WIRELESS NETWORK FOR WIRELESS NETWORK PERFORMANCE IN SUDANPERFORMANCE IN SUDAN

Presented By Presented By HUDA M. A. EL HAGHUDA M. A. EL HAG

University Of Khartoum – Faculty Of Mathematical University Of Khartoum – Faculty Of Mathematical SciencesSciences

hudaalhajj@yahoo.comhudaalhajj@yahoo.com

The Sudan General InformationThe Sudan General Information The capital city is Khartoum The capital city is Khartoum

34,475,690 (July 1999 ) estimated population34,475,690 (July 1999 ) estimated population

Area 967,494 sq mi (2,505,813 sq km), the largest Area 967,494 sq mi (2,505,813 sq km), the largest country in Africa, bordered by Egypt (N), the Red Sea country in Africa, bordered by Egypt (N), the Red Sea (NE), Eritrea and Ethiopia (E), Kenya, Uganda, and the (NE), Eritrea and Ethiopia (E), Kenya, Uganda, and the Democratic Republic of the Congo (S), the Central Democratic Republic of the Congo (S), the Central African Republic and Chad (W), and Libya (NW). African Republic and Chad (W), and Libya (NW).

The most notable geographical feature is the Nile The most notable geographical feature is the Nile River,700 kilometers across the country from the South River,700 kilometers across the country from the South to the north. to the north.

Rainfall in Sudan diminishes from south to north; thus Rainfall in Sudan diminishes from south to north; thus the southern part of the country is characterized by the southern part of the country is characterized by swampland and rain forest, the central region by swampland and rain forest, the central region by savanna and grassland, and the north by desert and savanna and grassland, and the north by desert and semi-desert.semi-desert.

The University of KhartoumThe University of Khartoum

TOPICSTOPICS

IntroductionIntroduction Why we need to analyze internet traffic in wireless links?Why we need to analyze internet traffic in wireless links? Transport protocol performance over wireless linksTransport protocol performance over wireless links What is self-similar traffic?What is self-similar traffic? Data Collection and MeasurementsData Collection and Measurements ConclusionsConclusions ReferencesReferences

IntroductionIntroduction

To properly model the performance of wireless data To properly model the performance of wireless data networks there must be a thorough understanding of the networks there must be a thorough understanding of the nature of internet traffic. Studies have shown that nature of internet traffic. Studies have shown that internet traffic is self-similar and heavy tailed in both local internet traffic is self-similar and heavy tailed in both local and wide area wired networks .and wide area wired networks .

Simulating this traffic cannot be done with Poisson Simulating this traffic cannot be done with Poisson models because these models result network designs models because these models result network designs which do not take into account the correct traffic which do not take into account the correct traffic behavior. The question is to determine whether wireless behavior. The question is to determine whether wireless data networks exhibit the same behavior. data networks exhibit the same behavior.

Why we need to analyze internet traffic in Why we need to analyze internet traffic in wireless links?wireless links?

Modeling assumptions affect our network designModeling assumptions affect our network design

For the fast-changing and heterogeneous Internet, For the fast-changing and heterogeneous Internet, determining the relevant model for a particular research determining the relevant model for a particular research question can be 95% of the work!question can be 95% of the work!

Users insist on having the same applications over Users insist on having the same applications over wireless links with the same quality of service that they wireless links with the same quality of service that they are getting over a wired linkare getting over a wired link

Transport Protocol Performance over Transport Protocol Performance over Wireless LinksWireless Links

Characteristics of wireless links that affect Characteristics of wireless links that affect transport protocol performancetransport protocol performance Packet loss due to corruption.Packet loss due to corruption. Delay variation due to link-layer error recovery, Delay variation due to link-layer error recovery,

handovers, and scheduling.handovers, and scheduling. Asymmetric and/or variable bandwidth (e.g., satellite).Asymmetric and/or variable bandwidth (e.g., satellite). Shared bandwidth (e.g., WIRELESS LANs).Shared bandwidth (e.g., WIRELESS LANs). Mobility.Mobility.

Self-Similar Data TrafficSelf-Similar Data Traffic

A phenomenon that is self-similar looks the same or A phenomenon that is self-similar looks the same or behaves the same when viewed at different degrees of behaves the same when viewed at different degrees of “magnification “ or different scales on a dimension .this “magnification “ or different scales on a dimension .this dimension can be space or time.dimension can be space or time.

Clusters are clusteredClusters are clustered

Queue sizes build up more than expected from Queue sizes build up more than expected from Poisson traffic.Poisson traffic.

Self similarity has a profound impact on performance Self similarity has a profound impact on performance

The higher the load on the networks, the higher the The higher the load on the networks, the higher the self-similarity.self-similarity.

If high levels of utilization are required, larger buffers are If high levels of utilization are required, larger buffers are needed for self similar traffic than would be predicted needed for self similar traffic than would be predicted based on classical queuing analysis. based on classical queuing analysis.

Self-Similar Data TrafficSelf-Similar Data Traffic

Data Collection and MeasurementsData Collection and Measurements

G round Station

Inte lsa t

Transm issionSystem

Internet Core G atew ay

Fram e Realy Netw ork

ISP's

Frame Relay LinkUpload/Download

Dv B/IP D ish

Dialup UsersDialup Users

Enterprise/SOHO

Fram e Relay LinkUpload/Download

UsersUsers

Dow

nloa

d O

nly

DvB/IP Dish

Download Only

Arabsa t

G round S tation

BT1

BT2

Emix

Teleg lobe

s1 /1s1 /3

s2/2

BT1BT2

Emix

MRTG (Multi Router Traffic Grapher) is the MRTG (Multi Router Traffic Grapher) is the software used for the collection of the datasoftware used for the collection of the data

Collects the traffic from the internet gateway Collects the traffic from the internet gateway routerrouter

time Units

1,8011,6011,4011,2011,0018016014012011

Tx

fram

es /t

ime

unit

1200

1000

800

600

400

200

0

time units

9018017016015014013012011011

Tx fr

ames

/tim

e un

it

1200

1000

800

600

400

200

0

time units

451401351301251201151101511

Tx

fram

es/ti

me

unit

1000

800

600

400

200

0

ssss

time units

2262011761511261017651261

Tx

fram

es/ti

me

unit 1000

800

600

400

200

0

Internet Statistics in Internet Statistics in Sudan Sudan

Traffic Behavior on Traffic Behavior on different Time scalesdifferent Time scales

Emix Upload Traffic Distribution

0

0.02

0.04

0.06

0.08

0.1

1 11 21 31 41 51 61 71 81

Utilization(%)

Pro

babi

lity

poisson

Observed

Emix Download Traffic Distribution

00.020.040.060.080.1

0.12

0 10 20 30 40 50 61 72 86

Utilization(%)

Pro

babi

lity

Poisson

Observed

BT1 Link Download Traffic Distribution

0

0.02

0.04

0.06

0.08

2 12 22 32 42 52 62 72 82 93

Utilization(%)

Pro

babi

lity

Poisson

Observed

BT1 Link Upload Traffic Distribution

0

0.02

0.04

0.06

0.08

2 12 22 32 42 52 62 72 82

Utilization (%)

Pro

babi

lity

Poisson

Observed

BT2 Link Download Traffic Distribution

0

0.02

0.04

0.06

0.08

0.1

3 13 23 33 43 53 63 73 83 93

Utilization(%)

Pro

babi

lity

Poisson

Observed

BT2 Link Upload Traffic Distribution

0

0.02

0.04

0.06

0.08

0.1

Utilization(%)P

roba

bilit

y

Poisson

Observed

Internet Internet Statistics in Statistics in Sudan Sudan

Traffic Traffic Distribution Distribution compared with compared with Poisson Poisson distributiondistribution

ConclusionsConclusions Modeling assumptions affect our network designModeling assumptions affect our network design

Internet traffic is self-similar and heavy tailed.Internet traffic is self-similar and heavy tailed.

Users insist on having the same applications over Users insist on having the same applications over wireless links with the same quality of service that they wireless links with the same quality of service that they are getting over a wired link.are getting over a wired link.

Wireless links affect transport protocol performance.Wireless links affect transport protocol performance.

ReferencesReferences A. Gurtov and S. Floyd, “A. Gurtov and S. Floyd, “Modeling Wireless Links for Modeling Wireless Links for

Transport ProtocolsTransport Protocols”, ”, November 2003November 2003.. D. Chandra, R.J. Harris, N. Shenoy, “D. Chandra, R.J. Harris, N. Shenoy, “Congestion and Congestion and

Corruption Loss Detection with Enhanced –TCPCorruption Loss Detection with Enhanced –TCP”” H. Balakrishnan, V. N. Padmanabhan, S. Seshan, R. H. H. Balakrishnan, V. N. Padmanabhan, S. Seshan, R. H.

Katz, Katz, “A Comparison of Mechanisms for Improving “A Comparison of Mechanisms for Improving TCP Performance over Wireless LinksTCP Performance over Wireless Links” ” IEEE\ ACM IEEE\ ACM Transactions on Networking (1996)Transactions on Networking (1996)

Hiba Mohammed Osman “Hiba Mohammed Osman “Internet Backbone Network Internet Backbone Network Traffic in SudanTraffic in Sudan” Masters Thesis” Masters Thesis

M. E. Crovella, "M. E. Crovella, "Self-Similarity in WWW Traffic: Self-Similarity in WWW Traffic: Evidence and Possible CausesEvidence and Possible Causes" " IEEE Trans. IEEE Trans. NetworkingNetworking, vol. 5, no. 6, Dec. 1997, pp. 835–45., vol. 5, no. 6, Dec. 1997, pp. 835–45.

ReferencesReferences S. Floyd and V. Paxson, “S. Floyd and V. Paxson, “Difficulties in Simulating the Difficulties in Simulating the

InternetInternet” , Transactions on Networking, August 2001.” , Transactions on Networking, August 2001. S. Floyd and E. Kohler, “S. Floyd and E. Kohler, “Internet Research Needs Internet Research Needs

Better ModelsBetter Models”, ”, HotNets-I, October 2002.HotNets-I, October 2002. William Stallings, “William Stallings, “High-Speed Networks and High-Speed Networks and

InternetsInternets”, First Edition, 1998 ”, First Edition, 1998 W. E Leland W. E Leland et al.et al., ", "On The Self-Similar Nature of On The Self-Similar Nature of

Ethernet TrafficEthernet Traffic," ," IEEE Trans. NetworkingIEEE Trans. Networking, vol. 2, no. 1, , vol. 2, no. 1, Feb. 1994, pp. 1–15.Feb. 1994, pp. 1–15.

THANK YOUTHANK YOU