网络通信流Multi-agent用于应用系统性能评价研究

Application Performance Evaluation by Multi-agent on Network Communication Flow

核心业务系统整体性能不佳可能导致系统在大业务量时出现功能、效率和安全性问题,因此系统的整体性能成为业界关注重点.现代的信息系统中新型软件架构层出不穷,软硬件和网络通讯之间的相互关联作用日益复杂,加大了系统性能瓶颈查找的难度.本文研究如何通过多智能代理技术等方法,截取、分析和回放网络数据流来测量软件性能.智能代理使得整个测试的各个分支具有通讯、协作的能力.依此模型方法,在两个大型业务系统进行了性能评价实测验证,证明此方法具有一定的优越性.

Core business information systems' performance problem may lead to the low business efficiency and system malfunction under pressure.Since many new software architecture occurred in modern information system,and the correlativity of software,hardware and network communication become more complexity,bottle-neck finding become quite difficult.This paper is about how to use multi-agent to do network stream analysis and replay to find the problem of system performance.This paper use agent technology to simulate the system's communication process.By capture segments and pattern reconstruction,many operations to test the performance could be simulated.Agents make the test components could communication and do some judgments.This paper put this method into practice to prove it.Practice was done on an electronic business platform and an exchange system.