Joint route-server load balancing in software defined networks using ant colony optimization

Abstract

Managing resource utilization can greatly improve system and network performance. Load balance (LB) is one important means for controlling resources. Traditional LB methods are usually inefficient, inflexible since there are no effective ways to obtain network statistics from each network device. This can be overcome by adopting the Software Defined Networking (SDN) approach. We present a dynamic LB for SDN, which aims to load balancing both the servers and the paths leading to the servers. To choose the best server, we use the least-loaded server LB policy. To select the best path to reach the chosen server, we adopt the Ant Colony System (ACS) algorithm, a successful variation of the nature-inspired ant colony optimization method. The proposed work therefore implements the joint ACS routing and dynamic server LB algorithm as a LB module in the SDN controller. This algorithm uses server-load and network statistics collected by the controller to find both the best server and the best path for network flows. As a proof of concept, the proposed algorithm is implemented as a module in the OpenDayLight controller, and experiments are carried over Mininet, an OpenFlow network emulator. The experimental results demonstrate that the proposed algorithm performs significantly well as compared to two other benchmark algorithms, achieving both higher network throughput and lower delay. This work would serve as an illustration of adopting SDN for system and network performance management, as well as a good example of using nature-inspired cooperative methods for SDN.