Hybrid Finite Automata-based Algorithm for Large Scale Regular Expression Matching

Fast data transmission put forward high requirements on network content matching (NCM). Due to the high time complexity, Nondeterministic Finite Automata (NFA) was unable to meet the demand of regular expression matching (REM) which was the core of NCM; Transfer NFA to Deterministic Finite Automaton (DFA) could enhance the throughput, but led to state explosion, which increased demand for memory. To balance memory and throughput, state explosion in the transformation from NFA to DFA has been analyzed and a new method DC-DFA is presented for large scale REM. DC-DFA is based on hybrid automata structure which composed of NFA and DFA. DC-DFA introduces GradeOne classification to cut the memory usage and deep classification to improve throughput. The results show that for serious state explosion, DC-DFA could reduce 75% DFA states and improve memory utilization efficiently while maintain high system throughput.