Handling high speed traffic measurement using network processors
Donghua Ruan,Chuang Lin,Zhen Chen,Jia Ni,Peter D. Ungsunan
DOI: https://doi.org/10.1109/ICCT.2006.341843
2006-01-01
Abstract:Traffic measurement is important in network management, accounting, worm detection, etc. It becomes more and more difficult today and cannot keep pace with the increasing speed of today's internet. Many new algorithms and hardware have been proposed to solve this problem. Network Processors, known for its nice tradeoff between performance and programming flexibility, are chosen as the platform. The goal of this paper is to find a suitable algorithm based on NPs to handle high speed network traffic measurement. Four algorithms, Raw measurement, Sampling, Multi-stage Filter and Multi-Resolution Space Code Bloom Filter (MRSCBF) are implemented and evaluated on Intel's IXP 2400 network processor. The results reveal that Sampling and Multi-stage Filters can fully exploit the parallel and heterogeneous architecture of network processor, so are suitable for high speed network traffic measurement on the network processor platform. MRSCBF, on the other hand, is not so efficient on network processors because of its complex process and frequent memory access.