基于多GPU的正则表达式匹配技术

Regular expression matching technology based multi-GPU

针对传统正则匹配性能低下的问题,设计了基于多GPU的正则表达式匹配引擎,并采用折半分组优化算法解决了有限状态自动机在大规模正则集合情况下由于空间爆炸无法使用的问题,并做了相关的优化,提升了数据匹配速度.实验结果表明:基于多GPU的正则表达式匹配性能较CPU提升了61倍,其数据吞吐率远优于其他加速方式.

In order to improve the performance of regular expression matching,the principles of regular expression matching was discussed,and the idea of making use of GPU（graphics processing unit） to accelerate matching was proposed.The regular expression matching engine based on multi-GPU was designed,and the problem that DFA（deterministic finite automaton）can not be used in largescale of regular expression set due to space explosion by binary algorithm was resolved.Experiment shows that the regular expression matching algorithm based on multi-GPU gets 61times speedup ratio than that based on CPU（central processing unit）,and its throughput is far superior to other accelerated manners.