CAJ | 학술논문

针对分布式报文分类算法内存消耗大、可扩展性差的问题,提出分布式元组空间叉积算法.该算法采用独立域搜索引擎与树状多级聚合网络的分类结构,在聚合节点使用计数型布鲁姆过滤器(CBF)加速搜索,利用剪枝技术降低CBF内存消耗.仿真结果表明,对于5×10~4条规模的9域规则库,聚合网络总内存消耗被控制在60Kb内,该算法的查找速度达到100Mp/s,且具有良好的可扩展性.
침대분포식보문분류산법내존소모대、가확전성차적문제,제출분포식원조공간차적산법.해산법채용독립역수색인경여수상다급취합망락적분류결구,재취합절점사용계수형포로모과려기(CBF)가속수색,이용전지기술강저CBF내존소모.방진결과표명,대우5×10~4조규모적9역규칙고,취합망락총내존소모피공제재60Kb내,해산법적사조속도체도100Mp/s,차구유량호적가확전성.
Aiming at the problem of huge memory consumption and poor scalability of distributed packet classification algorithm, this paper proposes distributed tuple space crossproducting algorithm. This algorithm uses a classification structure with independent field search engines and dendriform multilevel aggregation network, employs Counting Bloom Filter(CBF) in aggregation nodes to accelerate searching process, and utilizes pruning technology to reduce the memory consumption of CBF. Simulation results indicate that the total memory consumption of aggregation network is below 60 Kb when handling 9-field filter set of 5×10~4, the searching speed of this algorithm is 100 Mp/s, and the algorithm achieves better scalability.