Optimal virtualization of functionality for customer premise equipment

Network function virtualization has been introduced to extend the lifetime of customer premise equipment (CPE). Some functionalities of CPE for serving new service features have been expected to be shifted to a data center. However a lack of network cost analysis causes an argument of the worthiness of this shifting.We propose a determination framework for the functionality virtualization of CPE. Firstly, all functionalities of CPE are transformed to a graph structure. Secondly, the functionality virtualization is determined based on the functionality graph in order to minimize the network cost. The determination problem is formulated as a binary quadratic programming problem; however, searching for the optimal solution is NP-complete. Thus, we propose two heuristic algorithms (i.e., modified Karger's (mKar) and virtualization-cost-based determination (VCD)) to determine which functionality should be virtualized and placed in the data center. This paper performs both theoretical analysis and simulation to evaluate the performance of our algorithms. Those evaluations show that mKar is faster than VCD while VCD provides the solutions closer to the optimum than that of mKar. Conclusive remarks of our findings provide the significant insights into the way of functionality virtualization.