Project Topic: P2P Latency Distributions

1
Project TopicP2P Latency Distributions

• Michael K. Bradshaw

2
Motivation

• P2P systems are overlays
• Many attempt to minimize latency
• Three present testing methods
• Model count messages
• Simulate over fake network
• Measure over real network
• Need a fast way to study latency

3
Solution??

• Model the Systems latency as a cost distribution
• Less costly than measurement
• Comparable in different settings
• Problem
• Needs a distribution of latency costs
• Not presently available

4
Project

• Model the latency experienced between two
  randomly chosen peers as a distribution

5
Methodology

• Wide selection
• Obtain traces from different P2P systems
• Routers, filesharing, research
• Derive Distributions
• Classify P2P systems based on dist.

6
Suggestions

• Traces publicly available and accepted
• Pointers to distribution fitting?