Abstract: By analysis of state explosion in deterministic finite automata DFA, a novel algorithm Group2-DFA based on state constrains was proposed to reduce the memory usage. With the state constrains, states in NFA were classified into several groups. Group2-DFA introduces two-level classification and merges NFA and DFA together to a hybrid FA construction. The experiments show that Group2-DFA can reduce memory usage efficiently and keep high throughput with a small increase of memory reading time. With 300 regex rules, Group2-DFA can cut 75% states and achieve 1Gbps throughput.