High performance Deep Packet Inspection on multi-core platform

Abstract

Deep Packet Inspection (DPI) provides the ability to perform Quality of Service (QoS) and Intrusion Detection on network packets. But since the explosive growth of Intrnet, performance and scalability issues have been raised due to the gap between network and end-system speeds. This article describles how a desirable DPI system with multi-gigabits throughput and good scalability should be like by exploiting parallelism on Network Interface Card, network stack and user applications. Connection-based Parallelism, Affinity-based Scheduling and Lock-free Data Structure are the main technologies introduced to alleviate the performance and scalability issues. A common DPI application L7-Filter is used as an example to illustrate the applicaiton level parallelism.