DMCA
SCALABLE STRING MATCHING FRAMEWORK ENHANCED BY PATTERN CLUSTERING
| Citations: | 1 - 0 self |
BibTeX
@MISC{Xu_scalablestring,
author = {Bo Xu and Kai Zheng and Yibo Xue and Jun Li},
title = {SCALABLE STRING MATCHING FRAMEWORK ENHANCED BY PATTERN CLUSTERING},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
String matching plays an important role in content inspection based applications such as network intrusion detection/prevention and anti-virus. It is facing critical performance challenges due to the rapid increase in network bandwidth and the expansion in pattern set size. With multicore processors emerging as the dominant network processing platform, traditional one-dimension workload distribution model via flow-based traffic parallel processing can not fully exploit their computing power and cache hierarchy. In this paper, a scalable string matching framework is proposed by introducing another workload distribution dimension in pattern set. This framework distributes workloads in two dimensions: the network traffic dimension and the pattern set dimension. A novel pattern clustering mechanism named PCM is presented to optimize the pattern set partitioning. Experimental results show that the proposed framework obtains a throughput speedup of 60 % compared with the traditional one-dimension workload distribution mode on real-life rule sets while the PCM pattern clustering mechanism further improves the overall throughput by 15%~20%. The framework can adapt to various string matching algorithms and the PCM scheme can be applied to different leap-based algorithms.
Keyphrases
novel pattern flow-based traffic cache hierarchy real-life rule set scalable string network traffic dimension network intrusion detection prevention overall throughput different leap-based algorithm pcm pattern computing power workload distribution dimension throughput speedup content inspection multicore processor experimental result pcm scheme dominant network processing platform important role rapid increase traditional one-dimension workload distribution model critical performance challenge network bandwidth traditional one-dimension workload distribution mode
