Convergence time monitoring algorithm in hybrid software defined networks

Abstract

Network performance is extremely dependent on traffic monitoring. Therefore, Software Defined Network (SDN) technology is proposed to support the flow control and proper monitoring by providing a global view of the network. Unfortunately, replacing the entire traditional network to SDN is complex, which leads to the need of SDN switches deployment to the current network. Thus, a hybrid network environment has been emerged which consists of centralized controller, SDN switch and legacy routers. Hence, the advantage of the integration of traditional network and SDN have been taken place. The controller can collect SDN data instantly, while it waits for a long time to obtain the legacy network data. Consequently, failure detection and traffic management cannot be recognized in real-time. This research proposes a monitoring algorithm to monitor path state. It aims to reduce infrastructure cost in terms of replacing minimum number of legacy routers and minimizing the convergence time of collecting path load data. Significant paths are chosen by reconstructing load matrix using Singular Value Decomposition (SVD). SDN switches are then installed to cover these critical paths. As a result, critical paths can be directly addressed by the controller. On the other hand, the rest of the paths cannot be processed directly by the controller. Therefore, legacy path load data is estimated for the past time to support the controller for obtaining the current data. The proposed algorithm has been implemented over ISP topology of 24 nodes and 72 paths using Mininet emulator with Quagga routing software and Python-based open-source OpenFlow (POX) controller. The convergence time of the proposed algorithm takes only 12% more convergence time than the full SDN. Therefore, the proposed algorithm provides replacing one-third of legacy routers (8 out of 24) to SDN switches where the infrastructure does not need to be fully replaced which reduce the infrastructure cost.