回归分析作为一种估算资源消耗的方法常用于克服不易于直接测量的问题,但在多租户应用的动态负载环境下效果并不好。针对多租户环境下的动态负载变化,以CPU资源消耗为研究目标,提出一种基于回归分析方法的改良策略,减少共线性影响从而提高评估的精确度。实验结果表明,在动态的负载环境,所提方法能较好地适应多租户环境,平均估算误差得到改善。
As a method for estimating resource consumption,regression analysis is used to overcome the problem that uneasy to measure directly,but it does not perform well in the scenario of dynamic workload for multi-tenant application. This paper proposed a new regression analysis strategy to improve the accuracy of the estimation. This strategy considered the status of the workload and reduced the impact of its multicollinearity to improve the accuracy of the estimation. Test results show that,in continuously changing workload environment,the proposed strategy adapts to the multi-tenant environment well and the average estimation deviation has been reduced.