Optimal Routing for Competitive Service Providers in Network Virtualization Context
Achref El Amri, Aref Meddeb · 2024
In the context of Network Virtualization where service providers may instantiate virtual networks on a common physical infrastructure, multiple virtual servers may be instantiated on physical ones. These virtual servers are in different locations and provide the same service. Then, clients are served without knowing which server is replying and which path is crossed for their traffic routing. Service providers such as competitive ones who lease the physical infrastructure are responsible for adopting the suitable routing strategy. These providers aim to reduce the cost of leased resources while guaranteeing their clients' quality of service requirements. In this paper, we consider the optimal routing problem for competitive service providers in Network Virtualization context. We model it as a mixed integer linear program whose objective function is to minimize the cost under flow and servers' bandwidth constraints. We then solve the proposed program on a small instance of network topology. We notice that Traffic Concentration on the closest server yields an optimal solution: all the traffic of each client should be routed to the closest server in terms of hop count. We finally compare, through analytical models based on the M/D/1/N queue, the performance of Traffic Concentration with three load sharing techniques as a function of traffic intensity. We note that Traffic Concentration reduces the mean sojourn time for all the traffic intensities. However, the load sharing techniques outperform Traffic Concentration in terms of blocking probability for medium and heavy traffic intensities with comparable results for low traffic intensities.