云计算环境下云服务用户并发量的区间预测模型

发布时间：2018-03-31 21:30

本文选题：云服务　切入点：并发量　出处：《计算机学报》2017年02期

【摘要】：云计算环境下服务用户并发量的预测是云环境自适应资源调整的重要依据,但传统的单值预测所包含的信息量过少,受并发量不确定性影响明显,所以其不足以支持完备的自适应调整策略制定,因而会引发过多无效的调整动作.针对上述问题,该文提出一种云服务用户并发量区间预测模型,通过预测并发量的区间量化其不确定性.该模型利用梯度下降粒子群优化的支持向量机作为主要预测方法.为了更有效地预测不同类型的并发量,提出了一种基于自相关系数以及功率谱分析的AC-PS并发量特征判定规则,并针对不同特征并发量采取不同的区间构造方法.该文通过一个实例分析该区间预测模型对解决自适应资源调整无效问题的有效性,最后利用对比实验验证预测区间的准确性.结果表明,相对于其它方法文中提出的区间预测模型对各类并发量数据的预测精度均达92%以上,其预测效率有76.11%~96.15%的提升,因此提出的并发量区间预测方法能够为避免自适应资源调整无效问题提供可靠支撑.
[Abstract]:In cloud computing environment, the prediction of the amount of service user and output is an important basis for adaptive resource adjustment in cloud environment. However, the traditional single-valued prediction contains too little information and is obviously affected by the uncertainty of concurrent amount.Therefore, it is not enough to support the formulation of a complete adaptive adjustment strategy, which will lead to too many invalid adjustment actions.In order to solve the above problems, this paper presents an interval prediction model for cloud service users and traffic, which quantifies the uncertainty of concurrent quantities by interval quantization.The support vector machine (SVM) of gradient descent particle swarm optimization is used as the main prediction method in this model.In order to predict different types of concurrency more effectively, a new rule of AC-PS concurrency feature determination based on autocorrelation coefficient and power spectrum analysis is proposed, and different interval construction methods are adopted for different feature concurrency quantities.In this paper, an example is given to analyze the effectiveness of the interval prediction model in solving the invalid problem of adaptive resource adjustment. Finally, the accuracy of the prediction interval is verified by a comparative experiment.The results show that compared with other methods, the prediction accuracy of the interval prediction model for all kinds of concurrent data is over 92%, and the prediction efficiency is improved by 76.1196. 15%.Therefore, the proposed interval prediction method of concurrent variables can provide reliable support for avoiding the problem of invalid adaptive resource adjustment.
【作者单位】：东北大学计算机科学与工程学院;
【基金】：国家自然科学基金(61572116,61572117,61502089) 国家关键科技研发基金(2015BAH09F02) 省科技项目攻关项目(2015302002) 中央高校东北大学基本科研专项基金(N150408001,N150404009,N140406002)资助~~
【分类号】：TP393.09

【相似文献】