An Approach to Continuous Queries in Unstructured Overlays (Coquos)

Abstract

The current peer-to-peer (P2P) content distribution systems are running by their simple on-demand content discovery mechanism. The utility of these systems can be improved by incorporating two capabilities, first a mechanism through which peers can register their long term interests with the network so that they can be continuously notified of new data items, and second one is the peers to advertise their contents. Although researchers have proposed a few unstructured overlay-based publish-subscribe systems that provide the above capabilities, most of these systems require complex indexing and routing schemes, which not only make them highly complex but also gives less flexible on propagate the queries around transient peers. This paper argues that for many P2P applications, implementing full-fledged publish-subscribe systems is an complex task. For these applications, we study the alternate continuous query paradigm, which is a best-effort service providing the above two capabilities. we present a scalable and effective middleware, called CoQUOS stands for Continious queries in Unstructured Overlays for supporting continuous queries in unstructured overlay networks. Besides being independent of the overlay topology, CoQUOS preserves the simplicity and flexibility of the unstructured P2P network. Index Terms: Peer-to-peer networks, Continuous queries, Publish-subscribe systems, Random walk. I. Introduction: UNSTRUCTURED peer-to-peer (P2P)-based content/resource sharing platforms such as Gnutella and Kazaa have experienced excellent growths in the past days. The popularity of unstructured P2P networks can be based on the simplicity of their designs and their flexibility towards transient node population. Searching in these networks is essentially performed by circulating query messages in an ad-hoc manner and probing individual peer nodes. The current unstructured P2P content distribution systems suffer from certain serious limitations. One such limitation is their simple-on demand mechanism for content discovery. Peers in these systems discover data items by circulating queries within the overlay network. A peer receiving a query responds back to the initiating node if it has any matching content. Upon processing a query, the recipient node removes it from its local buffers1. Thus, a query expires after it completes its circulation within the network. In other words, the network forgets the queries once they have completed their circulation. For clarity purpose called this as a ad hock query model , and we refer to the queries as ad hoc queries. The ad hoc query model suffers from two main short comings First, an ad hoc query is only capable of searching and retrieving content that exists in the P2P network at the time the query was issued. In this scenario, the only way for a peer to discover newly added data items would be to repeatedly issue the same query. This is not desirable, since it creates unnecessary traffic within the network. Second, P2P systems that are purely based on the ad hoc query -model provide no support for peers to advertise or announce the data items they own to other interested peers. Advertisements are important for P2P applications where peers advertise their content. The above are shortcomings render the ad hoc query model .The many advanced P2P applications suffered from above two drawbacks. For handling the above situation we use a naive approach for tackling this problem would be to send advertisements to large subsets of peers through Flooding. However, this approach is unviable. When heavy messaging overheads, this scheme could gives the peers with unwanted advertisements. To address the above limitations we use well-studied paradigm of publish-subscribe (pub-sub) Systems is a possible approach to address the above limitations. The pub-sub involves intricate indexing and routing mechanisms . To overcome the above problem we present a scalable and effective middleware, called CoQUOS stands for Continious queries in Unstructured Overlays for supporting continuous queries in unstructured overlay. 1.1.) Paper Contributions: The diffculties in implementing pub-sub systems on top of unstructured overlays can be attributed to the inherent mismatch between the design requirements of pub-sub systems and the very nature of unstructured P2P systems. The pub-sub model provides weaker guarantees and individual implementations of may not provide and notification guarantees are weaker due to system failures or churn.