Time Series Analysis of P2P Networks to Improve Data Availability

In this paper we present a replication approach which utilizes the availability information of nodes in the network. By performing Box-Jenkins´s ARIMA model on the availability information, a prediction model is devised which helps to predict the future availability of the node. Based on this information highly requested data is replicated to nodes which complement each others availability in the network. This improves the overall availability and reduces the overlapping availability and reduces the excess storage for each replica. We have performed our evaluation on Kadamlia network. The Analysis showsour replication algorithm keeps data sustainable in the network. In our evaluation we show that life patterns along with the availability of nodes can be used to improve overall data availability. Our approach reduces over head compared to existing approaches to data availability.