关键词: 正则表达式匹配, 三态内容可寻址存储器, 并行字符索引, 分块存储, 低能耗

Abstract: Deep packet inspection (DPI) is a key function of network intrusion detection and prevention systems (NIDPS). TCAM-based regular expression matching algorithms have been proposed as a promising approach to improve processing speed, which is an important research direction of DPI. Ternary content addressable memory (TCAM) has the characters of high searching speed and small storage space, as well as the TCAM power consumption is proportionate to its storage space. Deterministic finite automaton (DFA) requires large storage space and the storage space of multi-stride DFA grows exponentially with the stride of DFA, which leads to high TCAM power consumption of DFA, especially for multi-stride DFA. This paper presents a parallel character-indexed multi-stride regular expression matching algorithm to address such limitation. This algorithm uses the idea of building parallel character indexes according to the stride of DFA, and reduces the number of activated TCAM blocks by using bitmap intersection, which in turn translates low TCAM power. Experimental results show that our algorithm reduces the TCAM power by more than 99.8% as well as the TCAM space usage by 48.5%~65.3%, and improves the matching throughput by 1.9~2.6 times compared with previous solutions based on multi-stride DFA.

Key words: regular expression matching, ternary content addressable memory (TCAM), parallel character index, block-based storage, low power