Dynamic Resource Prediction and Allocation in Clouds using Pattern Matching

Abstract

Objective: Cloud computing provides computational power as utility which can be acquired on demand. Cloud providers should have the ability to predict the future resource usage of an application. The prediction of resources aims to accommodate the dynamic resource requests generated by the applications and providing the resources when required. As the user’s demand changes over time, it becomes hard to maintain uniformity in delivering the service. Method: This paper proposes a dynamic technique that predicts resource usage in advance using simple patter matching technique. The prediction is based on the initial usage of the resources and executing tasks. The resources are offered as packages where each package has a bundle of processing capability. Initially the algorithm monitors the current usage and starts predicting and allocating resources after the usage pattern is available. Findings: The experiment displays initial failure and gradually increases in hit as prediction is more accurate with more available patterns. This way, there source usage can be predicted and necessary resources can be made ready in advance so that the user can get necessary services on demand. The prediction accuracy of the algorithm is 97.08%. The delay in initializing the resources is minimized by 25.381%. Application: The resource usage can be predicted and necessary resources can be made ready in advance so that the user can get necessary services on demand. Resource prediction can improve the quality of service parameter and cloud providers can be aware of the changing demands of the applications.