Optimal allocation approach of virtual servers in cloud computing

In this paper an optimal allocation approach of virtual servers to appropriate data centers is proposed for a network made up of collections of data centers and user groups. The virtual server allocation can be formalized as a linear programming (LP) problem which consists in minimizing the round-trip-times (latencies) between data centers and user groups. Our proposed allocation approach demonstrated good capabilities when evaluated under round-trip-time and group requirements (workload) changes. In compliance with the numerical results that are obtained using our proposed allocation approach, we observed that as long as the data center saturation is not achieved, increases in group requirements results in increases in the data center utilization percentage while the optimum doesn't change. However, when approaching the data center saturation our proposed approach ineluctably switches to another optimum resulting in a new reallocation of virtual servers. In contrast, by now varying round-trip-times between user groups and data centers neither data center utilization nor virtual server migration vary as long as the optimum remains unchanged. Nevertheless, beyond certain values of round-triptimes this optimum may obviously change inducing a virtual server re-allocation and variations in the data center utilization.