应用层协议识别是网络管理和安全中的重要功能，其处理速度已经成为制约高速网络设备发展的性能瓶颈之一。本研究针对千万兆网络，采用计算机体系结构方法，以高性能处理为目标，全面地研究适合应用层协议识别的软件方法与硬件体系结构。主要研究内容包括：高性能网络流维护方法与体系结构、基于网络流特征和协议特征的识别方法与体系结构、具备网络流维护的高性能应用层协议识别统一框架、基于隐私等级的应用层协议特征库建立等。我们拟设计一个应用层协议识别综合评价体系评估所提出方法和体系结构的适用性和有效性。在性能上，本研究采用软件模拟器和EDA工具进行初步分析，结合理论建模探讨解决方案的理论界限，并设计实现基于多核/引擎的软件原型系统和基于FPGA的硬件原型系统，测试实际效果。本研究内容将满足中高端网络设备在应用层协议识别方面未来5-10年的计算性能需求。

Application layer protocol identification is one of the important components and performance bottlenecks of network management and security systems. This research focuses on the bottleneck issues by presenting software methods and domain-specific architectures for 1 to 10 Gbps network, with performance as the main target. Our research plan includes methods and architectures for high performance flow table management, flow feature based and pattern based application protocol identification, the integrated framework to efficiently combine flow table management and protocol analysis. We also present the methods to setup privacy based protocol pattern set. To evaluate the applicability and effectiveness, we propose a full-scale evaluation model for all methods. For performance, we exploit software simulator, EDA tools and theoretical models to show the potential of our methods, and we design and implement multi-core/multi-engine based software prototype system and FPGA based hardware system for test and evaluation in real environment. Our research target is to provide methods and architectures to achieve enough performance for application layer protocol identification in next 5 to 10 years.