Single pattern search implementations in a cluster computing environment

Abstract

This paper provides an overview of different string matching algorithms in parallel environments. In this work, we have evaluated several algorithms, such as Knuth-Morris-Pratt, Boyer-Moore algorithm, Boyer Moore Horspool Algorithm, Zhu Takaoka algorithm, Quick Search Algorithm, BR Algorithm, Fast Searching algorithm, SSABS algorithm, TVSBS algorithm, ZTMBH algorithm and BRBMH algorithm. Static pattern of string matching algorithms are widely used for searching. For comparisons of speed different algorithms are implemented with parallel programming technique of MPI Standard, for a large text file with a pattern of small length. A Beowulf based homogenous cluster architecture was developed which has a performance of 81 Gigabits floating point operations per second in searching environment. The result of pattern searching showed the variation of performance with this customized cluster architecture for variable number of compute nodes with different pattern lengths. Nearly 80% of the searching code part works in parallel. The performance of string search algorithm is based on the network bandwidth and the selection of algorithms used. Depending on certain applications, different algorithm can be used. It is observed that the cluster searching architecture provides better speed with different algorithms. In general, from the experiments, BRBMH Algorithm is found to be efficient for a general string searching applications.