Multicriteria routing for guaranteed performance communications
Abstract
In this thesis, we investigate two routing problems. The first, which is known as the multiconstraint QoS (quality of service) routing problem, is to find a single path that satisfies multiple QoS constraints. For this problem, we consider two routing environments: (a) a given source node has detailed routing information provided by a link-state protocol, and (b) the source node has relatively simple routing information provided by a distance-vector protocol. First, we develop a greedy scheme, called MPLMR (multi-postpath-based lookahead multiconstraint routing), for case (a). MPLMR has an efficient “look-ahead” feature that uses the detailed information provided by link-state protocols. MPLMR has significantly better performance than competing schemes in the literature. We then develop a sequential path-search scheme, called SPMP (single-prepath multi-postpaths), for case (b). SPMP performs routing with simple routing information provided by distance-vector protocols, and maintains a small number of nodes involved in routing process. Hence, SPMP is suitable for multiconstraint QoS routing in the situations where reduction in computational/signaling overhead is a concern. The second problem that we deal with in this thesis is to find a minimum number of paths that can collectively satisfy constraints on channel demand, capacity, and survivability between a given pair of source and destination nodes in a WDM (wavelength division multiplexing) network. Different from previous survivable routing schemes for WDM networks, we introduce link failure probabilities to the problem. Because this routing problem is NP-hard, we develop heuristic multipath routing schemes: CPMR (conditional-penalization multipath routing) and SPMR (successive-penalization multipath routing). These schemes allow each link to be used for several channels. To deal with the difficulty that this link-sharing causes, we develop “link penalization” methods to control link-sharing. CPMR takes a long run-time to find a near-optimal solution, while SPMR uses a simple penalization method to reduce the run-time at the slight expense of the routing success rate. Via simulation, we show that our schemes achieve near-optimal routing success rates.
Degree
Ph.D.
Advisors
Siegel, Purdue University.
Subject Area
Electrical engineering|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.