Transparent Query Caching in Peer-to-Peer Overlay Networks

Peer-to-peer (p2p) systems such as Gnutella and KaZaa are routinely used by millions of people for sharing music and many other files over the Internet, and they account for a significant portion of the Internet traffic. The p2p traffic can be broken down into two categories: protocol messages for maintaining and searching the overlay p2p network, and data messages for downloading data files. This paper makes the following two contributions. First, we present a study of the locality in the collective Gnutella query protocol messages being initiated and forwarded by servents within the same organization. Second, we propose a transparent query caching scheme to reduce the bandwidth consumed by p2p traffic going in and out of the gateway of an organization. Our locality measurements show that there is significant locality in the collective queries going through a gateway forwarded by servents behind that gateway, and the locality increases with the population of those servents. Our proposed transparent caching scheme preserves the user experience, i.e., users continue to use the same servents as before, and queries will result in similar query hits with or without the caching running. Measurements of our transparent caching proxy in an experimental testbed of eight passive Gnutella servents in a LAN has shown a query cache hit rate of up to 38%, an uplink query traffic reduction of up to 40%, and a downlink query hit traffic reduction of up to 12% at the gateway.