基于马尔可夫过程模型的服务器集群数据可用性分析

Analysis of Cluster Servers′ Data Availability Using Markov Process Model

可用性是衡量服务器集群的重要指标.在基于马尔可夫过程的服务器状态转移模型基础上,推出根据单台服务器的可用性计算服务器集群可用性的形式化方法,建立了单服务器可用性、服务单元可用性和整个服务器集群可用性之间的量化关系,进而得到多副本分布集群的数据可用性公式.本文的成果有效支持给定服务结点可用性参数的约束下准确估计服务器集群的可用性,对于从可用性角度辅助用户进行集群设计具有理论价值和实际意义.

Accurate analysis and evaluation of availability has long been important in cluster servers＇ design. This paper presented a formal method to evaluate the availability of cluster servers accurately based on the Markov process model. The quantitative relationship among availability of clusters, availability of server, data replica distribution and the number of servers in clusters had been measured and a series of theory including a formula which can provide accurate evaluation of cluster serverst availability had been provided and proved. The contribution of this paper was not only in theory, but also in practice because the theories and conclusions can be taken into practice in cluster servers design to forecast the availability of system when the availability of server nodes were constrained.