EGOIST: Overlay Routing using Selsh Neighbor Selection - PowerPoint PPT Presentation

Title: EGOIST: Overlay Routing using Selsh Neighbor Selection


1
EGOIST Overlay Routing using Sel?sh Neighbor
Selection
Georgios Smaragdakis, Vassilis Lekakis, Nikos
Laoutaris, Azer Bestavros, John Byers and Mema
Roussopoulos
In Proc. of ACM CoNEXT'08
Presented by Hongfang Guo 2009-02-06
2
Motivation (1)

• Application of Overlay network
• Content distribution
• Peer-to-Peer file sharing
• Multimedia streaming
• Overlay routing
• Connectivity management
• A newcomer
• Rewire the links to deal with
• the churn

3
Motivation (2)

• Selfish behavior
• Has implications for connectivity management
• Operational purpose (bootstrapping)
• Seizing opportunities to maximize the local
  connection quality
• Harmful downsides
• Potential benefits from such selfish behavior
• Neighbor selection
• K-closest, K-Random,. (K is the number of
  neighbor)
• Selfish Neighbor Selection (SNS)

4
Motivation(3)

• Selfish Neighbor Selection

Using BR, a node chooses the best k neighbors
that optimize its connection quality to the
overlay,granted knowledge other nodes information
Using Best-Response(BR) neighbor selection
strategies.
10
Min.
Preference
5
The cost of shortest directed path between vi and
vj
Implications of Sel?sh Neighbor Selection in
Overlay Networks, INFOCOM07
5
Motivation(4)

• Background
• Selfish nodes selecting neighbors can providing
  global performance.
• One implication from the prior work
• Shortest-path overlay routing performs much
  better over SNS topologies than over random
• Build a SNS-inspired overlay ?
• Contributions
• EGOIST A prototype overlay routing network
• Implemented on PlanetLab
• Real measurement for evaluation
• Introduce a hybrid extension of SNS.
  Hybrid-EGOIST
• Vulnerability to malicious users

Theoretic
6
EGOIST(1)

• Basic Architecture
• Node listen to all control message of link state
  protocol
• Propagates them to immediate neighbors

133.3.3.3
X
144.4.4.4
122.2.2.2
111.1.1.1 122.2.2.2 25msecs 111.1.1.1
133.3.3.3 165msecs
111.1.1.1
7
EGOIST(2)

• Newcomers
• Connects to the system by querying a bootstrap
  node

133.3.3.3
Bootstrap node
A list of potential neighbors
122.2.2.2
144.4.4.4
W
Vi
111.1.1.1
99.9.9.9
8
Performance Evaluation

• Nodes
• 50 PlanetLab nodes for 2 months.
• Wiring policies
• EGOIST
• k-Random, k-Closest, k-Regular (DHT).
• Wiring frequency 60 seconds.
• Metrics of interest
• Delay (ping, Pyxida).
• CPU load (loadavg).
• Available Bandwidth (pathChirp).
• Control variables
• We vary the number (k) of neighbors.

11
7
30
1
1
9
Performance Evaluation(2)

• Performance of various neighbor selection

1.The result confirm the superiority of SNS
2.Out perform other policies especially k is
small
10
Performance Evaluation(3)
Performance Under Churn Hybrid-EGOIST k links,k1
for selfishly optimize its performance using SNS
and the k2k-k1 links to assure basic
connectivity under churn.
When K is small,donating 2 links simply cant
ensure the connectivity
1 nodes join or leave per second
11
Conclusion

• Problem is interesting
• The experiments are weak
• No novel solutions
• It is a paper different from others, for me, it
  is hard to read ,understand.

12
Thank You !
www.themegallery.com