Data Persistence in Structured P2P Networks with Redundancy Schemes

Data service is the principle application in structured P2P networks. And in many cases, we need the data to persist for a long time. So how to persist the data service and how long it can persist in the highly unstable P2P networks are interesting topics. To answer the first question, we have already got a number of redundancy schemes to compensate the loss of data caused by the churn of the networks. In this paper, we mainly focus on the second question: how long can data services persist with these redundancy schemes. We analyze the application model and generate several formulas to calculate the data persistence for different redundancy schemes using Markov chain. Then a performance evaluation among these redundancy schemes is conducted. We conclude that erasure coding scheme and the hybrid scheme get better data persistence than the replication scheme does in most cases except when the network churn rate is high and the average node availability is low. We also distinguish the temporary leave of a node from the permanent leave and demonstrate that data availability is also a crucial factor to determine the data persistence.