Abstract
Network Function Virtualization (NFV) replaces proprietary network appliances by softwarizing them in commodity servers and they are called as Virtual Network Functions(VNFs). As the network traffic changes dynamically, it is challenging to find the exact number of VNFs that can process the increase in demand. Moreover a VNF reconfiguration solution like migration or scaling has to be carefully chosen so that the increased traffic can be processed without violating the Quality of Service. We call this problem as VNF reconfiguration problem and provide an Integer Linear Programming formulation with implementation in CPLEX solver. From our simulations, we have concluded that our optimal formulation returns an efficient VNF embedding solution while reducing the reconfiguration costs. Scaling up VNFs are the most preferred choice but as the traffic increases significantly, scaling out VNFs are chosen more than other solution. Moreover our reconfiguration method has 30% less reconfiguration cost compared to other methods.