1 EL 933 Network Measurement and Traffic Engineering Yong Liu ECE Dept. Polytechnic University
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
EL 933Network Measurement and Traffic
Engineering
Yong LiuECE Dept.
Polytechnic University
2
About Myself
Graduated from UMass, Amherst, 02/2002 Joined ECE, Poly, on 03/15/2005 Research Area
networking: design/analysis/simulation computer systems: modeling/control/optimization
Teaching start from now…
Contact email: [email protected] web: http://eeweb.poly.edu/faculty/yongliu office: LC 252, ext. 3959
3
Internet Growth
Scale topology users speed
Application web/email p2p/overlay network game
Management understand control
0
20,000,000
40,000,000
60,000,000
80,000,000
100,000,000
120,000,000
140,000,000
160,000,000
1991
1992
1993
1994
1995
19961997
1998
1999
2000
2001
2002
source datawww.isc.comGNutella
KaZaa
Napster
Killer A
pplications !
4
Measurement & Traffic Engineering
Measurement characterization: traffic/topology/applications detection: failures/worms/DDoS prediction: delay/loss/throughput
• “Internet weather forecast”
Traffic Engineering (TE) network planning: topology design, link dimensioning routing: balance load across networks congestion control: achieve high resource utilization while
maintain network stability
5
Measurement Methodologiestaxonomy of approaches: active versus passive?
active: inject traffic, measure• tools: ping, traceroute,
passive: observe, measure existing traffic• tcpdump, netflow, snmp;
where measurements taken: network edge (host, servers) routers (“within” network) measurement boxes attached to links/routers
what metrics? delay, loss, rate, traffic type (http, p2p, udp, …) per-hop (local) or per-path (end-to-end)
6
IPMON @ Sprint ATL
7
Daily and Weekly Traffic Pattern
Traffic volume shows a clear diurnal pattern, with traffic tripling from06:00 through 12:00 noon EDT.
Traffic decreases by about 25% during the weekend. The two directions of the monitored link are not symmetric.
8
Counter-Strike Servers
Longitude histogram
9
Cooperative Association for InternetData Analysis (CAIDA)
AS Graph, April 2005 Router View
10
AT&T Backbone Topology
11
To be Covered … Traffic Analysis
traffic statistics: packets arrive according to Poisson? how user behavior affect traffic?
End-end path characterization estimate e2e delay, loss, throughput
Network Tomography from edge-based traffic measurements, infer internal link-
level loss, delay and utilization
Anomaly Detection DDos attacks/worm spreading/link failures
P2P Measurement P2P topology/user behavior/workload
12
To be Covered …
Traffic Matrix (TM) Estimation existent and new approaches
Optimal Routing optimize routes for single TM optimal routing in a changing world
Congestion Control TCP dynamic models Active Queue Management (AQM) closed loop analysis
Course Web: http://eeweb.poly.edu/el933/
13
Final Project Papers
IMC -- http://www.usenix.org/events/imc05/ PAM -- http://www.pam2005.org/ Sigcomm -- http://www.sigcomm.org/sigcomm2005/ Infocom -- http://www.ieee-infocom.org/2005/ Sigmetrics -- http://www.cse.cuhk.edu.hk/~sigm2005/ ICNP -- http://csr.bu.edu/icnp2005/
Reviews main contributions technical: assumptions, correctness/flaws what you like/dislike about the paper? possible improvements/extensions
30-45 minutes presentation, address questions
14
Questions & Comments?
?
Related Documents
