A Classification-Based Demand Trend Prediction Model in Cloud Computing
Cloud computing allows dynamic scaling of resources to users as needed. With the increasing demand for cloud service, a challenging problem is how to minimize cloud resource provisioning costs while meeting the user’s needs. This issue has been studied via predicting the resource demand in advance. Existing predicting approaches formulate cloud resource provisioning as a regression problem, and aim to achieve the minimal prediction error. However, the resource demand is often time-variant and highly unstable, the regression-based techniques can not achieve a good performance when the demand changes sharply. To cope with this problem, this paper proposes a framework of predicting the sharply changed demand of cloud resource to reduce the VM provisioning cost. In this framework, we first formulate the cloud resource demands prediction as a classification problem and then propose a robust prediction approach by combining Piecewise Linear Representation and Weighted Support Vector Machine techniques. Our proposed method can capture the sharply changed points in the highly unstable resource demand time series and improves the prediction performance while reducing the provisioning costs. Experimental evaluation on the IBM Smart Cloud Enterprise (SCE) trace data demonstrates the effectiveness of our proposed framework.
KeywordsCloud computing Capacity planning Piecewise Linear Representation Support Vector Machine
This work is supported by Natural Science Foundation of China under Grant No. 61503313 and the Jiangsu Key Laboratory of Image and Video Understanding for Social Safety (Nanjing University of Science and Technology), Grant No. 30920140122007.
- 4.Chang, C.C., Lin, C.J.: LIBSVM: a library for support vector machines. ACM Trans. Intell. Syst. Technol. (TIST) 2(3), 27 (2011)Google Scholar
- 5.Chatfield, C.: The Analysis of Time Series: An Introduction. CRC Press, Boca Raton (2013)Google Scholar
- 9.Iqbal, W., Dailey, M.N., Carrera, D.: Black-box approach to capacity identification for multi-tier applications hosted on virtualized platforms. In: 2011 International Conference on Cloud and Service Computing (CSC), pp. 111–117. IEEE (2011)Google Scholar
- 11.Jiang, Y., Perng, C.S., Li, T., Chang, R.: ASAP: a self-adaptive prediction system for instant cloud resource demand provisioning. In: 2011 IEEE 11th International Conference on Data Mining (ICDM), pp. 1104–1109. IEEE (2011)Google Scholar
- 13.Keogh, E., Chu, S., Hart, D., Pazzani, M.: An online algorithm for segmenting time series. In: Proceedings IEEE International Conference on Data Mining, ICDM 2001, pp. 289–296. IEEE (2001)Google Scholar
- 17.Whitle, P.: Hypothesis Testing in Time Series Analysis. Almqvist & Wiksells, Uppsala (1951)Google Scholar
- 19.Zhou, Q., Hong, W., Shao, G., Cai, W.: A new SVM-RFE approach towards ranking problem. In: IEEE International Conference on Intelligent Computing and Intelligent Systems, ICIS 2009, vol. 4, pp. 270–273. IEEE (2009)Google Scholar