Trying Linear Network Coding on a Network Flow Processor

Nowadays network coding has attracted a lot of attentions from researchers in fields of information and coding theory, wireless communications, P2P, etc, as it has enormous potential to improve network throughput and to utilize the maximum bandwidth in networks. Whereas due to its high computation complexity, it is not easy to implement network coding in a real network environment. Network Flow Processor (NFP) is a multi-core specific processor, widely used in switches, routers. NPF has a powerful computation capacity, since it is composed of 40 cores and each core can concurrently executes up to 8 threads. In this paper, we discuss about the network coding fundamental and the NFP architecture in order to make NFP support the encoding function of network coding. During the implementation of network coding on NFP, we use loop-based and table-based approaches for multiplication in finite field. With our experimental results, a single card of NFP can offer a throughput of encoding up to 43.9Mbps.