| Peer-to-peer networks are of great value for tapping the unused resources of computers residing at the edges of the internet. However, the recent rise in popularity of peer-to-peer applications has caused the sizes of their underlying networks to grow significantly. Because of their large size and decentralized nature, it is all but impossible to measure the efficiency of the protocols driving these networks in the real world. To facilitate the evaluation of peer-to-peer protocols prior to deployment in real networks, this thesis presents a scalable simulation architecture which can be extended to support arbitrary peer-to-peer protocols. In particular, the simulator models the competition of messages for bandwidth as traffic increases. For validation purposes, an implementation of multiple versions of the Gnutella protocol is presented. It demonstrates the ability of the simulator to model arbitrary peer behaviour and handle heterogeneous protocols. |
