Date Awarded


Document Type


Degree Name

Doctor of Philosophy (Ph.D.)


Computer Science


Rahul Simha


Broadband high-speed networks, such as B-ISDN, are expected to play a dominant role in the future of networking due to their capability to service a variety of traffic types with very different bandwidth requirements such as video, voice and data. to increase network efficiency in B-ISDN and other such connection oriented networks, the concept of a virtual path (VP) has been proposed and studied in the literature. A VP is a permanent or semi-permanent reservation of capacity between two nodes. Using VPs can potentially reduce call setup delays, simplify hardware, provide quality of service performance guarantees, and reduce disruption in the event of link or node failure.;In order to use VPs efficiently, two problems must be solved. With the objective of optimizing network performance, (1) the VPs must be placed within the network, and (2) network link capacity must be divided among the VPs. Most previous work aimed at solving these problems has focused on one problem in isolation of the other. at the same time, previous research efforts that have considered the joint solution of these problems have considered only restricted cases. In addition, these efforts have not explicitly considered the benefits of sharing bandwidth among VPs in the network.;We present a heuristic solution method for the joint problem of virtual path distribution and capacity allocation without many of the limitations found in previous studies. Our solution method considers the joint bandwidth allocation and VP placement problem and explicitly considers the benefits of shared bandwidth. We demonstrate that our algorithm out-performs previous algorithms in cases where network resources are limited. Because our algorithm provides shared bandwidth, solutions found by our algorithm will have a lower setup probability than a network that does not use VPs as well as a lower loss probability than provided by VPDBA solutions produced by previous algorithms. In addition, our algorithm provides fairness not found in solutions produced by other algorithms by guaranteeing that some service will be provided to each source-destination pair within the network.



© The Author