Predictive Resource Allocation Scheme with Pattern Matching

Abstract

Cloud provides huge processing power, immense memory and great storage for its customers on demand but a Cloud setup never have enough amount of resources actually. To accomplish the basic rule of Cloud game, always provide resources as customer required, Cloud play tricky games. Cloud system, keep eye on resource usage and maintain a fragile balance among supply and demand. The availability of resources, elasticity and quality of services increase the profit and monetary benefits of Cloud owner. The following treatise presented a simple but power technique that enables a Cloud system setup its balance of supply and demand with less extra efforts. This is a simple pattern matching method that provide the prediction of resource usage over a time period. In this method, system setup groups of resources and then deploy these resources initially. Later the usage pattern extracted from each group and prediction of next utilization of this bundle on the base of KNN classifier. The output was perpetual in its nature because system have no initial information to predict but every course of execution it increased its ability to get closer of actual requirements. The final results showed the maximum up to 98.5% and reduce the latency as well.