Performance evaluation of cloud service considering fault recovery
Performance evaluation of cloud service considering fault recovery
复制标题
DOI：
10.1007/s11227-011-0551-2
复制
发表时间：
2011-02
期刊：
影响因子：
--
通讯作者：
Bo Yang;Feng Tan;Yuan-Shun Dai
中科院分区：
文献类型：
--
作者：
Bo Yang;Feng Tan;Yuan-Shun Dai
云计算是IT领域的一个新趋势，引起了人们的广泛关注。在云计算中，服务可靠性和服务性能是两个重要的问题。为了提高云服务可靠性，可以使用诸如故障恢复的容错技术，这反过来又影响云服务性能。这种影响值得深入研究。虽然有一些云/网格服务的可靠性和性能的研究，很少有人解决的问题，故障恢复及其对服务性能的影响。本文对考虑故障恢复的云服务性能评估进行了详细的研究。我们认为恢复处理节点和通信链路。放宽了通常采用的用户服务请求泊松到达的假设，服务请求到达间隔时间可以采用任意概率分布。子任务的优先约束也被考虑。推导了服务响应时间的概率分布，并给出了数值算例。本文提出的云计算性能评价模型和方法能够得到符合实际的评价结果，对云计算相关决策具有一定的实用价值。
Cloud computing is a recent trend in IT, which has attracted lots of attention. In cloud computing, service reliability and service performance are two important issues. To improve cloud service reliability, fault tolerance techniques such as fault recovery may be used, which in turn has impact on cloud service performance. Such impact deserves detailed research. Although there exist some researches on cloud/grid service reliability and performance, very few of them addressed the issues of fault recovery and its impact on service performance. In this paper, we conduct detailed research on performance evaluation of cloud service considering fault recovery. We consider recovery on both processing nodes and communication links. The commonly adopted assumption of Poisson arrivals of users’ service requests is relaxed, and the interarrival times of service requests can take arbitrary probability distribution. The precedence constraints of subtasks are also considered. The probability distribution of service response time is derived, and a numerical example is presented. The proposed cloud performance evaluation models and methods could yield results which are realistic, and thus are of practical value for related decision-makings in cloud computing.