A Churn-Resistant Peer-to-Peer Web Caching System

Prakash Linga, Indranil Gupta, and Ken Birman
SSRS '03: Proceedings of the 2003 ACM workshop on Survivable and self-regenerative systems

Overview

Kelips is a system that employs probabilistic schemes and a self-regenerating data structure. The load of Kelips is bounded under node arrival and departure. Dissemination in Kelips occurs through gossiping, the dissemination latency can be decreased by increasing background overhead. Using Kelips for web cache sharing would lead to rapid lookups and low overheads, even under high churn rates.

When a user requests an object the local cache is first checked for a copy. If there is no copy in the local cache, it will try to locate the object in the Kelips system. If the object cannot be found in the Kelips system, the origin web server is requested for the object.