基于免疫机理的计算系统性能监控和评估

对于分布式高性能计算系统来说,模拟免疫机理实现计算系统的性能监控和评估是一个崭新的研究途径.分析和比较了免疫机理和计算系统抗衰之间的异同,构建了基于多Agent的系统抗衰逻辑模型,模拟免疫机理对计算系统的性能进行监控、诊断和建立性能衰退的数学模型,并在仿真实验中评价了性能监控对于所监控的计算节点的影响.在此基础上以一个音像资源事务处理系统为背景进行了应用研究,给出了一个两阶段超指数分布的数学模型来评估性能.仿真实验和应用研究的结果表明方法是有效可行的.

An Approach to Immunity-Based Performance Monitoring and Evaluation for Computing Systems

Simulating biological immune mechanism to implement performance monitoring and evaluation of computing systems is a new research approach for the distributed high performance computing environment. First, to simulate the biological immune mechanisms, the immune process's characteristics and computing system's rejuvenation are analyzed and compared. Second, to monitor and diagnose performance degradation, the logical and mathematical models of system rejuvenation are set up, which use an extension of the standard?proc performance interface offered by Linux systems to collect resource information form both local and remote hosts, then the principle of self-nonself discrimination inspired by immunology to diagnose system performance. On the basis of both the logical and mathematical model, an application regarding audio-video resource transaction process system as background is studied, a two-stage hyper-exponential model is proposed to evaluate system performance, and the effect imposed by a performance monitoring agent is evaluated. The results of experiments and case study indicate that this method is effective and feasible to monitor and evaluate the performance of a distributed computing system.