EPITELLA: Improving the Gnutella Search Algorithm through Epidemic Spreading Models for Complex Networks

Search algorithms in unstructured P2P networks such as Gnutella use flooding-based techniques for communication and, as a consequence, they produce high message overhead. More dynamic algorithms such as Gnutella’s Dynamic Query Protocol take into account the user’s desired number of results and network topology properties to increase scalability. However, these algorithms only work well for popular files and often fail to locate rare content.

In this paper, we introduce EPITELLA, an enhanced Gnutella which uses a search algorithm based on epidemic-style information dissemination techniques, that shows good performance in finding both rare and popular files. It exploits the structure of the underlying network in order to maximise its search horizon and minimise the number of needed search messages. The presented simulation results show that EPITELLA is more exhaustive in finding rare files and produces less overhead in finding popular files, compared to Gnutella.