基于贝叶斯模型的云服务服务质量预测

Cloud service QoS prediction method based on Bayesian model

针对如何分配一个未来一段时间内满足QoS要求的云服务和感知可能将要发生的QoS违规的问题,提出一种基于时间序列预测方法的云服务QoS预测方法。该预测方法利用改进的贝叶斯常均值(IBCM)模型,能够准确地预测云服务未来一段时间内的QoS状态。实验通过搭建Hadoop集群模拟云平台并收集了响应时间和吞吐量两种QoS属性的数据作为预测对象,实验结果表明:相比自回归积分滑动平均(ARIMA)模型和贝叶斯常均值折扣模型等时间序列预测方法,基于改进的贝叶斯常均值模型的云服务QoS预测方法的平方和误差(SSE)、平均绝对误差(MAE)、均方误差(MSE)和和平均绝对百分比误差(MAPE)均比前两者小一个数量级,因此具有更高的预测精度;同时预测结果对比图说明提出的预测方法具有更好的拟合效果。

For Quality of Service（ QoS） guarantee of cloud service areas,a cloud service QoS prediction method based on time series prediction was proposed to select an appropriate cloud service which met QoS requirements of cloud user and perceive QoS violation may occur. The improved Bayesian constant mean model was used to predict QoS of cloud service accurately. In the experiment,a Hadoop system was established to simulate cloud computing and a lot of QoS data of response time and throughput were collected as predicted object. The experimental result shows that compared with Bayesian constant mean discount model and Autoregressive Integrated Moving Average（ ARIMA） model,the proposed prediction method based on improved Bayesian constant mean model is one order of magnitude smaller than the previous methods in Square Sum Error（ SSE）,Mean Absolute Error（ MAE）,Mean Squared Error（ MSE） and Mean Absolut Percentage Error（ MAPE）,so it has higher accuracy; and the comparison of prediction accuracy illustrate that the proposed method also has better fitting effect.