Capacity and load-aware software-defined network controller placement in heterogeneous environments

Relying upon a single centralized controller in Software-Defined Networks may lead to scalability problems. To make the network scalable and keep latency low for wide area networks, using multiple controllers is proposed. The network is partitioned into some domains, each one is monitored by a controller. However, the number of required controllers, their location and load balancing among them, to prevent controller overloading, have become major challenges in the distributed control plane. Controller Placement Problem (CPP) is widely known as a solution to tackle these issues. Nonetheless, different costs and capacities of controllers should also be considered in solving CPP. In this paper, the problem is formulated as a location-allocation model and our proposed framework solves it in two phases. Regarding the different costs of deployment, types and capacities of controllers in the market, the first phase focuses on determining the required number of controllers while minimizing the total cost. Using the result, the second phase is to solve the location-allocation problem to balance the controller load with our introduced fair load distribution function and to reduce inter-controller latency. Two greedy procedures for location and also allocation are designed for our proposed framework algorithms to solve the models and numerical results show their efficiency.