Cooperative Data Center Selection for Optimal Service Performance: An ILP Formulation

Cross layer optimization is well-known to offer certain cost and performance advantages when designing and operating data networks [1]. It is reasonable to anticipate similar advantages when applying such optimizations across the network layer and the application layer too. In this paper, the authors define one optimization problem whereby the service-to-hosting data center mapping is performed taking into account the available resources in both the data center (CPU cycles) and network (bandwidth) subsystems. The objective function of the optimization problem accounts for the amount of both CPU cycles and bandwidth, which are assigned to each requested service, with the assumption that the user´s Quality of Experience (QoE) strictly depends only on these two factors. The problem is formulated using ILP and its numerical results are discussed to gain insights into the potential QoE gains that might result from using this cross layer (cooperative) resource assignment technique.