Date of Award

1991

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Computer Science

First Advisor

Sitharama S. Iyengar

Abstract

There exist at least two models of parallel computing, namely, shared-memory and message-passing. This research addresses problems in both these types of systems, and proposes efficient parallel (Shared-Memory Model) and distributed (message-passing) algorithms for a variety of graph related computational problems. In part I, we design algorithms for three generic problems in distributed systems: set manipulation, network structure recognition and facility placement. We present optimal distributed algorithms for recognizing rectangular-mesh networks. The time and message complexity of our algorithm is linear in the number of nodes in the network. We also lay the foundation for the recognition of 2-reducible, outer-planar and cactus graphs. These algorithms have a message complexity of O(kn), where, k is the number of isolated two degree nodes in the network. We introduce the problem of reliable r-domination and design unified optimal distributed algorithms for the total, reliable and independent r-domination on trees. The time and message complexity of our algorithm is O(n), where n is the number of nodes in the tree. In the domain of set manipulation we design optimal algorithms for determining the intersection of sets in a distributed environment, where each processor is assumed to have its own set. The time and message complexity of our set intersection algorithm is O(mn), where m is the cardinality of the smallest set. In part II of our research we design optimal algorithms for r-domination and efficient parallel algorithms for the p-center problems on trees. We also present an optimal algorithm for computing the maximum independent set on intervals i the EREW-PRAM model. The r-domination problem on trees can now be solved in O(logn)time with O(n/logn) processors using the EREW-PRAM model. A parallel algorithm for range searching is developed using the concept of distributed data structures. We show that O(logn) search time can be effected for a range search on n 3-dimensional points using (2.log$\sp2n-14.logn + 8$) processors. Our algorithm can easily be generalized for the case of d-dimensional range search. (Abstract shortened with permission of author.).

Pages

167

Share

COinS