Paper #: 22
Title: Tracing a large-scale Peer to Peer System:  an hour in the life of 
Gnutella


Problem:
1.  This paper studying traffic patterns generated by Gnutella.
2.  This paper suggests a caching scheme to improve Gnutella searches.




The authors found that there was a great deal of redundancy in the queries made.
They suggest an alternative metric to hit rate for measuring the success of a 
query.

Findings:  

The author found that:

1. Gnutella traffic is very bursty.
2. Gnutella traffic is roughly consistent across the overlay network
3. Nine out of ten searchs return no results.


Idea:  
The authors put forward a caching scheme that takes into account the source of a 
query, and the sources of the cached queries.  The scheme suggests returning 
query requests by combining cached queries, with queries made selectively to 
nodes that have not been queried yet for the request.  This scheme greatly 
reduces the traffic on the overlay network, and increases the response time for 
queries.



Implementation:
Traffic patterns where studied by placing modified Gnutella client on three 
separate continents, and then studying the queries received, and forwarded by 
the these clients.