Dynamic placement of proxies in optimal locations for router assisted hierarchical approach and hierarchical reliable multicast network

Abstract

The high performance computing models cloud, grid and utility computing are the different forms of distributed computing that allow users to tap into vast network resources to complete their work. This thesis primarily focuses on scalability and reliability issues of multicast transport layer protocols. Though User Datagram Protocol (UDP) is an accepted multicast transport layer service, there is no guarantee that a given packet reached all recipients in the multicast group. Three different analyses are made and performance measures are discussed. The proxies in the LMS are static, selected at the beginning of multicast session. HRM network employs routers as proxies that consolidate feedbacks and perform retransmission scheme. The algorithms k-median problem, Facility location problem, Tree-based Algorithm, Filter placement problem, and Greedy algorithm perceive solutions for proxy placement issue. The recent studies conclude that greedy based algorithm performs the best to achieve optimal solution. This thesis proposes setup and refreshes functions to minimize the link transmissions to implement Maximum Reduction Scheme (MRS) algorithm Greedy-MRS. This thesis proposed a distributed greedy based algorithm which minimizes overall bandwidth utilization within a group, and keeps minimum information in router for the selection of proxy node in HRM network. The algorithm introduces a new transmission reduction step that finds a new proxy which provides maximum reduction in traffic. In Greedy-MRS, the source node repeatedly communicate with its children to find set of proxies where distributed greedy algorithm on proxy node gather data from its children only. The results also indicate that bandwidth utilization of proxy selection process reduces by 50% when compared to Greedy-MRS algorithm. Since the proposed algorithm distributes the proxy selection process to subgroup, the complexity of placing a single proxy is O(h) where h is the height of subtree. newline newline newline