Parallel Matching of Regular Expressions with BSP Automata

Abstract

Regular expression matching is a core component of many applications including patterns search in text, deep inspection of packet or lexical analysis. Sequential regular expression matching lacks efficiency for large amount of data whereas parallel regular expression matching overhead requires a large number of processors to become negligible. This paper presents a transformation from regular expression (RE) into a parallel form named BSP regular expression (BSPRE). This transformation added to the transformation from BSPRE to parallel automata (BSPA) enable the parallel matching of regular expression. We compare this approach to enumeration method and observe substantial improvement for small number of processors. The automatic transformation from RE to BSPA through BSPRE is the first example of an infinite family of BSP programs that can be generated automatically and that are not simple specializations of a finite library.