Power Optimization of Large Scale Mobile Cloud System Using Cooperative Cloudlets

Reducing the total power consumption and network delay are among the most interesting issues facing large scale Mobile Cloud Computing (MCC) systems and their ability to satisfy the Service Level Agreement (SLA). Such systems utilize cloudlet based infrastructure to support off-loading some of user's computationally heavy tasks to the cloudlets. However, the limited capabilities of the cloudlet system (in terms of the ability of serve different request type and the ability to serve users in large geographical regions) represent serious challenges to achieve those objectives. To cover the users demand for different types of services and in wide geographical regions, cloudlets cooperate among each others by passing user requests from one cloudlets to another. By adapting this cooperation, the total power consumption per request will be increased so that it includes the power consumption between the user and the local cloudlet and the power consumption of passing the request to a remote cloudlet. In this paper, we consider two types of cloudlets: local cloudlets and global cloudlets. The global cloudlets are a special kind of local cloudlets but with higher capabilities. The user can connect only to the local cloudlet and sends all its traffics to it. If the local cloudlet cannot serve the desired request, then the request is moved to other local cloudlet. If no local cloudlet can serve the request, then it is moved to a global cloudlet in which it can serve all service types. We optimize the power consumption for large scale cooperative cloudlets and evaluate the proposed model under two realistic scenarios. The result prove that the proposed model can be used to optimize power consumption in large scale MCC systems.