Private:pCDN:Peer Matching
From NMSL
Revision as of 12:13, 25 April 2008 by Nitin.chiluka (talk | contribs)
some intro
Proposed Algorithm
Ideas
- Verification of AS Shortest Paths
- Experimental Setup
- # Peers: {100, 300, 500, 1000, 3000, 5000, 8000}
- # Senders: {2, 5, 10, 15, 20}
- # Sessions: 1000
- Algos: Random, Network, Geo, AS_Order(ISP-Friendly)
- Metrics: Min, Avg, Max, Output
- Output: Hops (Aggregate, P2P, C2P)
- Search Efficiency:
- If multiple shortest paths are existing, choose the least "cost" (s2s=0, p2p=1, c2p=2)
- Check practicality
- What kind of data we need initially and periodically
- How often we need them
- We implement the algorithm to determine AS relationships
- Large graphs
- Can we get the US & CA graph?
- Can we optimize this matrix.
- Usage of ISP + Network / Geolocation Algorithms
- We reduce the number of IP hops
- Diameter of the graph
References and Links
- links to various resources
