Abstract:

Among many aspects of routing in peer-to-peer overlay networks we pay attention to the following ones.

• Performance. The latency between two peers should not exceed a certain bound even for large networks.
• Scalability. The performance must not degrade rapidly when the number of peers grows.
• Dependability. A message must reach the destination even with high churn, connectivity failures or congestion in the underlying network.
• Security. Routing protocols should not present security vulnerabilities and must be robust against attacks.

This talk introduces a concept of cyclic routing in P2P overlay networks. Our research is at the beginning but we believe it could lead to novel routing mechanisms with the properties listed above. To support this opinion, we describe several applications of cyclic P2P routing and discuss possible implementations.

In a P2P overlay network, a message is routed through a multi-hop path. For each hop a node LOCALLY selects the next hop. Local routing decisions can be inefficient, espessially when the network is large. Information on the global state is provided by restricting topology to a simple structure such as a ring (structured P2P overlays). This information, however, can be insufficient to achieve routing with desired properties. Therefore, the topology could be enhanced with global structures other than links between neigboring nodes. The nodes can also collect information about the network state beyond their neighbors. In our approach, cycles present bidirectional paths available in an overlay network. The global topology can be enhanced by collecting global cycles; nodes can share them in routing. On the other hand, each node can collect cycles itself to select optimal routes to forward the messages.