云计算中服务质量的概率预测和评估方法研究

A Study of Probabilistic Prediction and Evaluation for Quality of Web Service in Cloud Computing

基于连续时间马尔可夫链提出了在云计算中组合服务质量的预测模型,模型针对QoS属性定义了基于线性时序逻辑区间的质量约束规范,进一步使用M/M/1队列实现QoS概率预测。另外,引入了Va R(Value-at-Risk),以货币的形式度量服务违例风险。并定义了不同网络环境下的补偿策略,给出了组合云服务聚合的Va R计算方法。通过将QoS预测模型在智能电网中实例测试,得到实际预测时间小于服务更新时间,预测准确率较高。风险评估机制与其他计算方法相比可信度较高。

Firstly,prediction model of composite service quality is proposed in cloud computing which is based on continuous time Markov chain. In the model, constraint specification for QoS properties is defined based on Linear-time Temporal Logic, and QoS probabilistic prediction is further realized by using M/M/1 queuing. In addition,VaR （ Value-at-Risk） evaluates violation risk in monetary terms. It defines compensation policies in different networks, and proposes calculation method of aggregation VaR. Through the practical test in the smart grids, prediction spent less time comparable with the updating rate usually considered for smart grids, and had the advantage of high accuracy. Comparing with alternative calculation methods, the mechanism of risk assess-ment is efficient and high in reliability.