面向下一代互联网实验平台的新型报文处理模型——EasySwitch

A Novel Packet Processing Model for Next-Generation Internet Experimental Platform——EasySwitch

下一代互联网实验平台应能提供网络原型系统快速开发及部署能力,以有效支撑新型互联网体系结构关键技术的实现和验证.基于FPGA(Field Programmable Gate Array)技术构建的网络实验平台可以提供较高的可编程性和性能,而它对硬件逻辑设计能力的要求则严重限制了平台的广泛应用.新型报文处理模型EasySwitch通过优化设计并预置通用报文交换及处理逻辑,有效实现用户定制报文处理逻辑与通用报文处理逻辑解耦;良定义的用户模块接口则使用户仅需关注业务特定逻辑实现,有效简化用户逻辑设计.理论分析表明,EasySwitch可通过提供确定性资源约束模型,有效支持FPGA资源的优化利用.此外,EasySwitch具有较低的报文调度处理延迟,对实验系统输入流量真实特性影响较小.EasySwitch模型在NetMagic平台的有效实现及应用表明该模型可为下一代互联网新型报文处理机制及协议的快速设计、开发和验证提供有力支撑.

To support the implementation and verification of the key technologies of the next-generation Internet architecture,the corresponding network experimental platforms should provide the ability of rapid development and deployment of network prototype systems.The FPGA（Field Programmable Gate Array） based experimental platforms provide both high programmability and high performance.However,the requirement of the hardware logic design ability greatly limits the widespread application of these platforms.A novel packet processing model,EasySwitch,is proposed to reduce the barriers of the network experimental system design and development.By pre-placing the common packet switching and processing functions,EasySwitch decouples the customized packet processing logic and the common packet processing logic.Its well-defined standard User Module socket interfaces allow the users to focus on implementing their own logic,simplifying the programming of hardware logic.Theoretical analysis shows that EasySwitch can optimize the hardware resource utilization of the FPGA with its deterministic resource constraint model.Moreover,with the low packet scheduling latency,EasySwitch model has less impact on the real characteristics of the input traffic from the experimental system.The implementation of EasySwitch model on the NetMagic platform demonstrates that EasySwitch can provide efficient support for the rapid design,development and verification of the novel packet processing mechanisms and network protocols.