Storage and Availability Aware Fragment Placement for P2P Storage Systems

Cloud storage systems are of great importance today due to features such as availability, mobility and performance. In addition to this, peer-to-peer systems ensure fault tolerance and scalability. The ideal combination of them is a P2P cloud storage system with an autonomous, self-configuring and fault-tolerant network. Besides performance, incentives are a key aspect of such P2P storage systems, with money being the most commonly used one. This paper puts forward the idea of a shared interest in data as the only incentive in order to build a viable P2P storage solution for both personal and enterprise storage. However participation in such a system is voluntary, making replication critical to ensuring data availability in the face of node churn. Erasure coding is a famous way of increasing data durability while maintaining a low expansion factor. The distribution of these encoded fragments though, across multiple peers is an opportunity for optimization which dictates much of the system performance. The paper proposes a fragment placement strategy to optimize the usage of peer resources and ensure symmetry in the system. For the purpose of node lookup and resource discovery in the network, an implementation of Kademlia distributed hash table is used.