Replication is a popular technique for increasing data availability and improving perfor- mance in peer-to-peer systems. Maintaining freshness of replicated data is challenging due to the high cost of update management. While updates have been studied in structured networks, they have been neglected in unstructured networks. We therefore confront the problem of maintaining fresh replicas of data in unstructured peer-to-peer networks. We propose techniques that leverage path replication to support efficient lazy updates and provide freshness for cached data in these systems using only local knowledge. In addition, we show that locally available information may be used to provide additional guarantees of freshness at an acceptable cost to performance. Through performance simulations based on both synthetic and real-world workloads from big data environments, we demonstrate the effectiveness of our approach.

Replication is a popular technique for increasing data availability and improving perfor- mance in peer-to-peer systems. Maintaining freshness of replicated data is challenging due to the high cost of update management. While updates have been studied in structured networks, they have been neglected in unstructured networks. We therefore confront the problem of maintaining fresh replicas of data in unstructured peer-to-peer networks. We propose techniques that leverage path replication to support efficient lazy updates and provide freshness for cached data in these systems using only local knowledge. In addition, we show that locally available information may be used to provide additional guarantees of freshness at an acceptable cost to performance. Through performance simulations based on both synthetic and real-world workloads from big data environments, we demonstrate the effectiveness of our approach.