An Intelligent SLA based Cloudlet Allocation Strategy using Machine Learning

The internet has widened its horizon far beyond the web browser. The service oriented technologies like Live Streaming videos over Netflix, Hulu, YouTube make up to 60% of today's internet traffic. These services require massive computation power; cloud computing which is onset conveyance of computational power, database storage, and other IT related services through a cloud platform via internet on pay-as-you-go pricing model serves the required purpose. However, the computational power available at the data centers is often limited and should be utilized efficiently so that the multiple request can be handled simultaneously and the terms and conditions reached between the client and the Cloud Service Provider are honored. In this paper, Multilevel Feedback Queue is used for load balancing followed by an SLA based resource allocation i.e. Virtual Machines to the requests i.e. cloudlets using the A* algorithm. The A* derives its values from the Ant Colony Optimization and Heuristic search. Together they form a Machine Learning technique that eventually learns the path that should be followed by a particular size of cloudlets. Thus, decreases the execution time of the cloudlets while necessarily maintaining the Quality of Service.