| With the development of the Internet technology, the huge population of the Internet users drives the Internet economy growth rapidly in China. However, network bandwidth and the IP addresses have a great demand gap, which are aroused by the behindhand infrastructural construction, limits the development pace of the Internet. Traditional Internet infrastructure is generally designed by specific hardware solutions to achieve the high throughput with high costs and poor scalability, so difficult to satisfy the fast-changing networking protocol. In the other side, the software solutions on the general-purpose hardware provide a low cost, highly flexible, expandable method to build the Internet infrastructure, but with the low performance of thoughput. As performance of the high compute-intensive graphics chip GPU enhances, GPU is not just used to improve image processing and graphics rendering, but also used to apply in the calculation of large-scale data processing. With the advantage of the powerful parallel processing capability of GPU, the heterogeneous computing platform applying in the Internet infrastructure provides a feasible idea to achieve high thoughput.This paper presents a model of the parallel processing network packets based on the heterogeneous computing platform. In this model, the Internet infrastructure is abstracted into three Layers: I/O Layer, CPU Scheduling Layer, and GPU Parallel Processing Layer, then an embodiment of the model is given by using the technology of the high performance network packets I/O engine, Multi-threading, and the Open CL heterogeneous computing. Finally IPv4/IPv6 translator is applied into the model of the parallel processing network packets, which interworking between IPv6 network and IPv4 network. It proves that the model works to improve the thoughput.The model can be widely used in the network infrastructure with its high versality, which will promote the network infrastructure construction and the IPv6 transition in the future. |
PDF Full Text Request