HyperCuP - Shaping Up Peer-to-Peer Networks

Peer-to-peer networks are envisioned to be deployed for a wide range of applications. However, P2P networks evolving in an unorganized manner suffer from serious scalability problems, limiting the number of nodes in the network, creating network overload and pushing search times to unacceptable limits. We address these problems by imposing a deterministic shape on P2P networks: We propose a graph topology which allows for very efficient broadcast and search, and we describe a broadcast algorithm that exploits the topology to reach all nodes in the network with the minimum amount of messages possible. We provide an efficient topology construction and maintenance algorithm which, crucial to symmetric peer-to-peer networks, does not require any central server nor super nodes in the network. Nodes can join and leave the self-organizing network at any time, and the network is resilient against failure. Peer-to-peer networks are envisioned to find a broad range of applications, moving way beyond their current application as infrastructure for file sharing and exchange such as in Napster or Morpheus (2). However, current P2P approaches face unresolved problems. Genuine P2P networks such as Gnutella, consisting entirely of identical peers in terms of their role in the network, do not scale to a large number of nodes due to their use of inefficient search mechanisms relying on undirected broadcast. Other P2P approaches such as Napster are deprived of the fundamental advantage of P2P, namely the strict peer symmetry, by employing central servers. Super-peer networks such as Morpheus share this disadvantage. In this paper, we focus on creating a deterministic network structure which can then be exploited to carry out efficient searching and broadcasting. We achieve this by organizing peers in a P2P network into a graph structure based on hypercubes in our system HyperCuP (Hypercube P2P). We present a distributed algorithm which is capable of maintaining the graph structure efficiently. In its current implementation, our approach is especially suitable for P2P networks in which desired information is widely distributed in the network, spread among a larger number of nodes. Instead of routing point queries, a query would be broadcasted along the topology, trying to pick up as