Traffic engineering approach to virtual-link provisioning in software-defined ISP networks

Abstract

-In this paper, we propose a new approach to virtual-link provisioning in ISP (Internet Service Provider) networks. The approach relies on Software Defined Networking (SDN) architecture based on OpenFlow that allows highly granular traffic splitting across multiple paths. The traffic control logic of SDN controller is divided into offline and online component. The online component handles dynamic arrivals of virtual-link requests in accordance with Service Level Agreements (SLAs) requirements. The offline component is responsible for periodic optimization of traffic distribution in the network. In this way, we tend to increase acceptance ratio for virtual-link requests and minimize degradation of best-effort traffic in a scalable manner. Our simulation results show that the proposed method for virtuallink provisioning outperforms solutions that rely on Constrained Shortest Path First (CSPF) and Equal Cost Multi-Path (ECMP) routing.

Dynamic optimization of load-balancing and reconfiguration overhead in SD-ISP networks

“…In order to evaluate the performance of the proposed approach, we built a simulator in Python that uses CPLEX [14] to solve the optimization problems. We used POP (Points of Presence)-level Sprint ISP topology [6], where link capacities were set to 1000 units, while delay on each link was derived from a great circle distance between the connected POPs and the speed of light in the fiber. Equal link capacities were assumed based on [15].…”

“…The pseudo-code of the heuristic is given in Algorithm 1. We start by solving the linear programming (LP) optimization model that minimizes MLU for the current traffic matrix [6]. The model can be solved in polynomial time if delay-matching tunnels for both traffic classes are known in advance.…”

“…The emergence of Software Defined Networking (SDN) redefined the traditional network architecture by separating the control plane from the data plane. This created opportunity to implement more sophisticated TE techniques, since SDN controller is able to continuously track the network state and reconfigure the data-plane dynamically [1][2][3][4][5][6][7]. In order to ensure an optimal network operation and meet QoS requirements, SDN controller must respond to the network events (e.g.…”

“…Thus, there is a key problem to balance between the routing optimality and the reconfiguration frequency. The conventional way to cope with the above problem is to perform throughput optimizations periodically (offline) and use greedy QoSaware routing algorithm for online path computation [6,7]. Increasing the period between TE optimizations means better reliability and stability, but the network keeps a suboptimal configuration longer.…”