Dynamic resource allocation exploiting mobility prediction in mobile edge computing

In 5G mobile networks, computing and communication converge into a single concept. This convergence leads to introduction of Mobile Edge Computing, where computing resources are distributed at the edge of mobile network, i.e., in base stations. This approach significantly reduces delay for computation of tasks offloaded from users' devices to cloud and reduces load of backhaul. However, due to users' mobility, optimal allocation of the computational resources at the base stations might change over time. The computational resources are allocated in a form of Virtual Machines (VM), which emulate a given computer system. User's mobility can be solved by VM migration, i.e., transfer of VM from one base station to another. Another option is to find a new communication path for exchange of data between the VM and the user. In this paper we propose an algorithm enabling flexible selection of communication path together with VM placement. To handle dynamicity of the system, we exploit prediction of users' movement. The prediction is used for dynamic VM placement and to find the most suitable communication path according to expected users' movement. Comparing to state of the art approaches, the proposal leads to reduction of the task offloading delay between 10% and 66% while energy consumed by user's equipment is kept at similar level. The proposed algorithm also enables higher arrival rate of the offloading requirements.