摘要
To work out a solution for answering the question when it is suitable for emergency response strategies to be executed, an emergency algorithm for enhanced survivability of a key service is presented. First, based on the central limit theorem and the hypothesis testing theory, the confidence interval of each key service＇s history average service response time in the host server and the spare server can be figured out, respectively. This can also be updated dynamically by a proposed method using the method of time slide window. Then, according to the five kinds of distributed situations of the current service response time＇ s confidence interval in the host server and the spare server, the proposed algorithm can dynamically choose the appropriate emergency policies such as resource reconfiguration, service degradation or service drifting, etc. in fight time. Thus, the key service request can be finished within its expected deadline by users as far as possible. Furthermore, the whole process of dynamic configuration is transparent to users. Finally, simulation tests are performed to prove the feasibility.
为了解决"何时对应急服务进行响应是合适的?"这一问题,提出了一种增强关键服务可生存性的应急算法.首先,基于中心极限定理和假设检验理论,分别计算求得各关键服务在主、备服务器上的历史平均服务响应时间的置信区间,并且提出了一种基于时间滑动窗口的置信区间动态更新的方法;然后,根据主、备服务器上服务响应时间置信区间的5种分布情况,提出的算法可以为应急服务在不同的时间响应区选择合适的应急策略,如资源重配、服务降级或者服务迁移等,以此保证关键服务请求能够在用户期望的服务截止期内完成,且整个配置过程对用户透明.最后,仿真试验表明了该算法的可行性.
基金
Specialized Research Fund for the Doctoral Program of Higher Education (No20050217007)
the Weaponry Equipment Foundation of PLA Equipment Ministry (No51416060104CB0101)
