Towards an enhanced revenue model for virtual networks embedding problem

This paper investigates the impact of incorporating memory usage in the revenue computing objective function defined in [1]. Towards this end, an enhanced revenue function has been proposed and tested on both live data, obtained through the CoMon monitoring facility for PlanetLab [2], and GT-ITM based simulations [3]. It has been found that, under the condition of uniform and constant weights, the addition of a memory usage term significantly increases the net revenue generated and also modifies its temporal characteristics. Hence, neglecting memory usage will produce flawed resource allocations and thus inefficient virtual network embedding systems.