Controlled Update Propagation in Peer-to-Peer Networks

Title: Controlled Update Propagation in Peer-to-Peer Networks

Abstract: This research proposes a new architecture called Controlled Update Propagation (CUP) for peer-to-peer networks. CUP aims to improve the performance of these networks by asynchronously building caches of index entries while answering search queries and propagating updates of index entries to maintain these caches. The architecture is designed to reduce query latency, network load, and the cost of updating caches. It also gives peer-to-peer nodes the flexibility to use their own incentive-based policies for receiving and propagating updates. The small propagation overhead incurred by CUP is more than compensated for by its savings in cache misses.

Research Question: How can we design an architecture that efficiently manages caches and updates in peer-to-peer networks to improve query latency, network load, and update cost?

Methodology: The CUP architecture is based on two logical channels per neighbor: a query channel and an update channel. Query channels are used to forward search queries for items of interest to the neighbor closest to the authority node for those items. Update channels are used to asynchronously forward query responses (first-time updates) and update index entries cached at a neighbor. Queries travel "up" the query channels, while updates travel "down" the update channels.

Results: CUP reduces the average miss latency by as much as a factor of three under unfavorable conditions and by more than a factor of ten under favorable conditions. It refreshes intermediate caches, reduces query latency, and reduces network load by coalescing bursts of queries for the same item.

Implications: The CUP architecture provides a comprehensive solution to the problem of maintaining caches and updating index entries in peer-to-peer networks. It improves the performance of these networks by reducing query latency, network load, and update cost. Additionally, it gives peer-to-peer nodes the flexibility to use their own incentive-based policies for receiving and propagating updates.

Link to Article: https://arxiv.org/abs/0202008v1 Authors: arXiv ID: 0202008v1