Resource Allocation Heuristics for Network Virtualization

Abstract

Network Virtualization allows overcoming the limitations of the current Internet. It enables the embedding of virtual networks on physical ones. Service providers share the same physical infrastructure in order to provide their own services and deploy their new ones. The allocation of available physical resources is an important issue for Network Virtualization. Due to the hardness of this problem, many heuristics are proposed to simplify its resolution. In this paper, we present two novel heuristics for physical resource allocation which are Traffic Concentration and Load Sharing. With Traffic Concentration heuristic, service providers concentrate all the traffic on the shortest path. However, with Load Sharing heuristic, they share the load on multiple paths. We compare the performances of these heuristics in terms of latency, jitter, packet loss ratio, connection to server probability, node and link utilization, throughput and leased bandwidth as a function of the number of used servers. Simulation results demonstrate that the choice of resource utilization has an impact on the performance: with Traffic Concentration, we gain in terms of cost of leased bandwidth and jitter. But, with Load Sharing, we gain in terms of latency, packet loss ratio, connection to server probability and throughput.