Many applications including one-to-many file transfer, information dissemination (e.g., stock quotes and web cache updates), distance learning, shared whiteboards, multi-party games, and distributed computing can benefit from reliable multicast. The goal of this dissertation is to design and evaluate large scale multicast loss recovery architectures and protocols for IP multicast capable networks, that make efficient use of both the network and end-system resources and scale to applications that can have thousands of receivers spanning wide area networks. One of the important problems in multicast loss recovery is that of a receiver receiving unwanted retransmissions of packets lost at other receivers. We present a new approach to scoping retransmissions in which a single multicast channel is used for the original transmission of packets and separate multicast channels are used for scoping retransmissions to “interested” receivers only. We find that a small to moderate number of multicast channels can be recycled to achieve almost perfect retransmission scoping for a wide range of system parameters. We also propose two mechanisms for implementing retransmission channels, one using multiple IP multicast groups and the other using a single IP multicast group in conjunction with additional router support. We show that the second approach reduces both host processing costs and network bandwidth usage. Another problem arises when the sender alone bears the burden of handling loss-feedback and supplying retransmissions to a large group of receivers. Local recovery approaches, in which entities other than the sender aid in loss recovery, distribute the loss recovery burden and also reduce network bandwidth consumption and recovery latency. We propose a new local recovery approach that co-locates designated repair servers with routers at strategic locations inside the network. We demonstrate the superior performance of our approach over traditional approaches. We address the important issues of repair server placement, repair server resource requirements, and performance degradation due to insufficient resources. We also demonstrate how the repair server functionality can be provided as a dynamically invocable/revocable service with minimal router support.
Identifer | oai:union.ndltd.org:UMASS/oai:scholarworks.umass.edu:dissertations-3258 |
Date | 01 January 1999 |
Creators | Kasera, Sneha Kumar |
Publisher | ScholarWorks@UMass Amherst |
Source Sets | University of Massachusetts, Amherst |
Language | English |
Detected Language | English |
Type | text |
Source | Doctoral Dissertations Available from Proquest |
Page generated in 0.0014 seconds