针对云服务器中存在软件老化现象,将造成系统性能衰退与可靠性下降问题,借鉴剩余使用寿命(Remaining useful life,RUL)概念,提出基于支持向量和高斯函数拟合(Support vectors and Gaussian function fitting,SVs-GFF)的老化预测方法.首先,提取云服务器老化数据的统计特征指标,并采用支持向量回归(Support vector regression,SVR)对统计特征指标进行数据稀疏化处理,得到支持向量(Support vectors,SVs)序列数据;然后,建立基于密度聚类的高斯函数拟合(Gaussian function fitting,GFF)模型,对不同核函数下的支持向量序列数据进行老化曲线拟合,并采用Fréchet距离优化算法选取最优老化曲线;最后,基于最优老化曲线,评估系统到达老化阈值前的RUL,以预测系统何时发生老化.在OpenStack云服务器4个老化数据集上的实验结果表明,基于RUL和SVs-GFF的云服务器老化预测方法与传统预测方法相比,具有更高的预测精度和更快的收敛速度.
Abstract
Aiming at the problem that software aging in cloud servers will cause system performance degradation and reliability descending,a software aging prediction method based on support vectors(SVs)and Gaussian func-tion fitting(SVs-GFF)with the use of the concept of remaining useful life(RUL)is proposed.Firstly,the statistic-al characteristic indexes of aging data on a cloud server are extracted,and then support vector regression(SVR)is used to sparse the data of statistical characteristic indexes into support vector sequences.Then,the Gaussian func-tion fitting(GFF)model based on density clustering is established to fit the aging curves of support vector se-quence data under different kernel functions,and the Fréchet distance optimization algorithm is used to select the optimal aging curve.Finally,based on the optimal aging curve,the remaining useful life before the system reaches the aging threshold is evaluated to predict when software aging occurs.The experiment results on four aging data sets of an OpenStack cloud server show that,the proposed cloud server aging prediction method based on remain-ing useful life and SVs-GFF has higher accuracy and faster convergence speed compared with traditional prediction methods.
关键词
云服务器/软件老化/支持向量回归/高斯函数拟合/剩余使用寿命
Key words
Cloud server/software aging/support vector regression(SVR)/Gaussian function fitting(GFF)/re-maining useful life(RUL)
维普
万方数据