Virtual resource prediction in cloud environment: A Bayesian approach. (April 2016)

• Record Type: Journal Article
• Authors: Shyam, Gopal Kirshna; Manvi, Sunilkumar S.
• Is Part Of: Journal of network and computer applications. Volume 65(2016)
• Page Start: 144
• Page End: 154
• Publication Date: 2016-04
• Subjects: Bayesian network -- Resource prediction -- Cloud computing; Microcomputers -- Periodicals; Computer networks -- Periodicals; Application software -- Periodicals; Micro-ordinateurs -- Périodiques; Réseaux d'ordinateurs -- Périodiques; Logiciels d'application -- Périodiques; Application software; Computer networks; Microcomputers; Periodicals; 004.05; 004
• DOI: 10.1016/j.jnca.2016.03.002
• Languages: English
• ISSNs: 1084-8045

Abstract

Abstract: With increase in requirement for dynamic execution of user׳s application in cloud, resource prediction techniques are gaining a lot of importance as the foundation for online capacity planning and virtualized resource management in data centers. There is a wide scope for the development of accurate resource requirement prediction methods to ensure that the virtualized resources do not suffer from over or under-utilization. We propose a Bayesian model to determine short and long-term virtual resource requirement of the CPU/memory intensive applications on the basis of workload patterns at several data centers in the cloud during several time intervals. However, the model is applied to predict resource(s) of all applications in general. The parameters considered for prediction in the model are day of week, time-interval of application access, workload, benchmarks, and availability of virtual machines etc. The model is simulated by using the SamIam Bayesian network simulator and workload traces of Amazon EC2 and Google CE data centers in dynamic scenarios. The performance is evaluated by considering benchmarks of CPU intensive applications (web based). The proposed model is able to predict virtual resources in a cloud environment with better accuracy as compared to other models.