A Content-Based Publish/Subscribe over Two-Tier DHT Utilizing Domain Ontology

Several design alternatives have been advocated for content-based publish subscribe communication infrastructure over structured peer to peer networks. These efforts have contributed significantly towards efficient implementations of subscription specification, installation of publications and subscriptions and notification delivery. There are some design issues which need to be addressed. These issues include semantic expressiveness of subscriptions, loss of notifications under high churn and the amount of routing traffic generated for installation of publications and subscriptions. In this paper, we propose a content-based publish/subscribe framework over DHT (Distributed Hash Table) based peer to peer networks to address above design issues. We have utilized the heterogeneity of nodes in terms of computing resources and uptime to construct a two-tier DHT. The nodes in the upper tier are termed as super nodes that have high computing resources and better uptime. Only these super nodes are used to store subscription and publication metadata and act as rendezvous nodes for them. We construct a domain ontology based concept neighborhood overlay of these rendezvous nodes. Further, these nodes are also used to cache popular contents and metadata for improved lookup efficiency. Our proposal has limited effect of high churn conditions as rendezvous nodes have considerably higher uptime. We demonstrate our scheme based on an ontology used by ACM for classification of its publications. We compared our approach with Ferry, a popular content-based publish subscribe framework over DHT with respect to performance under churn, routing traffic and notification efficiency. Simulation results indicate that our scheme performs well compared to Ferry in these aspects.