Flexible update management in peer-to-peer database systems

Abstract

Promising the combination of dynamic configuration, scalability and redundancy, peer-to-peer (P2P) networks have garnered tremendous interest lately. Before long, this interest extended to P2P sharing of disparate databases. While many have explored the data placement and coordination difficulties involved in such systems, far fewer have examined the unique challenges involved in updating P2P-distributed data items. In addressing that problem, we present an adapted version of the traditional quorum-consensus approach that is better suited to the P2P domain. Further, we explore the many tradeoffs introduced by this flexible update scheme, especially in the areas of data freshness, data availability, access latency and redundancy. Our simulation results demonstrate that if the data replication or quorum levels are sufficiently high it is possible to achieve a near-zero probability of stale data access. However, since more nodes are involved in quorum-building, the cost of high data confidence is slower access times. Keeping with the flexibility theme, our philosophy is to leave control in the hands of individual peers, letting them choose parameters and tradeoff performance to meet their unique needs.