Peer-to-peer cached data distribution network as an alternative for the client-server model

Abstract

For over three decades, the client server architecture has dominated the software development industry as the go-to solution of building a web based solution. Even with the drawbacks such that it possess such as, minimum bandwidth from client to server as the maximum bandwidth and single source of failure, the client server model is still considered the most viable solution due to its simplistic approach and relative ease of scaling. The main aim of this dissertation is to provide a new architectural strategy to compete with the client server architecture based on the peer to peer data distribution model proposed by the BitTorrent algorithm in conjunction with the browser based peer connection protocol of WebRTC. This study is focused on proposing a new architecture to overcome the limitation of client server architecture while incorporating modifications to the basic BitTorrent algorithm to provide a more suitable approach to finding peers in the fast shifting environment of websites. This thesis will discuss on how the BitTorrent algorithm‟s peer selection can be altered by introduction of return response time for better peer identification and how to maintain fairness for all parties in this environment. This research looks into how the performance of the web solution can be improved with minimal additional external components while maintaining standard protocols to provide a cache based data distribution as an alternative for the client server architecture. In this research the positive impact of adding the return response time for better selection of peers and the fairness calculations considering the upload and download, to remove the biased peer selection in BitTorrent will be taken into consideration comparing to the improvement of speed and download rates with respect to the client server architecture.