Elastic Scheduling of Virtual Machines in Cloudlet Networks

This paper addresses the problem of elastic virtual machine scheduling in cloudlet networks, and the objective is to maximize the dominant elasticity of computing and bandwidth resources in cloudlet networks. First, a mathematical formulation of the problem is presented. Second, the bandwidth consumption incurred by infra-job communication is studied. Next, we focus on the optimal scheduling of input jobs to minimize dominant load of cloudlet networks under the constraint of minimum bandwidth consumption, which is an NP-Complete problem. In that case, we design a 2-approximation algorithm. Evaluations results validate the efficiency of our algorithm.