Packet Classification Algorithm, Based On Data From The Multi-core Network Processor

Since the late 80s of last century, the Internet has transformed from an early day low speed network connecting predominantly educational institutions to a gargantuan fast growing commercial infrastructure. Various network services continue to emerge. Traditional best-effort network equipment can not meet the requirements. Network equipment needs to provide enhanced services such as admission control, resource reservation, QoS, VPN and other differential services.The key mechanism, that enables differential services, is the packet classification, which was a hotspot back then. Scholars have proposed several classic packet classification algorithms. Earlier classic packet classification algorithms are mostly based on mainstream network equipment of that era, and some are based on specific hardware. With the maturity of multi-core network processor, packet classification algorithm based on multi-core network processor becomes a very important research topic.This paper first introduces researches based on multi-core network processor, and problems caused by multi-core parallel processing. It then describes the concept of packet classification, introduces several common packet classification algorithms and presents a summary of these algorithms. Based on these, this paper describes the problem of packet classification based on multi-core network processor and several solutions, including multi-queuePower formula used to evaluate packet classification performance based on multi-core network processor, Tuple Space based and Jump-bit based parallel strategy, and HS-ewhich is an improved HyperSplit. At the last, this paper presents results of several experiments which show thatthese solutions can improve the performance of packet classification under certain conditions.