Scalable flow-based network processor for premium network services

Abstract

This paper presents OmniFlow system that is suitable for future Internet and premium network services requiring multi-layer protocol processing. It is implemented with multi-core processors running a flow-based packet processing for high parallelism performance. It provides for shortening development time, cutting down system cost, and scalability. Its key technology is a flow-based network processor that contains 40 RISC cores, being implemented as ASIC. The network processor is the 40Gbps packet forwarding programmable engine to process incoming packets through two SPI-4.2s.