Section: New Results

Optimizing rules placement in OpenFlow networks: trading routing for better efficiency

Contributors: Chadi Barakat, Xuan Nam Nguyen, Damien Saucez and Thierry Turletti

The idea behind Software Defined Networking (SDN) is to conceive the network as one programmable entity rather than a set of devices to manually configure, and OpenFlow meets this objective. In OpenFlow, a centralized programmable controller installs forwarding rules onto switches to implement policies. However, this flexibility comes at the expense of extra overhead as the number of rules might exceed the memory capacity of switches, which raises the question of how to place most profitable rules on board. Solutions proposed so far strictly impose paths to be followed inside the network. We advocate instead that we can relax routing requirements within the network to concentrate on the final destination to which the traffic should forwarded, not how to route to this destination. In [19] we illustrate the concept, with an optimization problem that gets the maximum amount of traffic delivered according to policies and the actual dimensioning of the network. The traffic that cannot be accommodated is forwarded to the controller that has the capacity to process it further. [19] also demonstrates that our approach permits a better utilization of scarce resources in the network. We extended the work by stating that in many situations (e.g., data-center networks), the exact path followed by packets has not significant impact on performances as long as packets are delivered to their final destination decided by the endpoint policy. It is thus possible to deviate part of the traffic to alternative paths so as to better use network resources without violating the endpoint policy. In [20] , we propose a linear optimization model of the rule allocation problem in resource constrained OpenFlow networks with loose routing policies. We show that the general problem is NP-hard and propose a polynomial time heuristic, called OFFICER, that aims at maximizing the amount of carried traffic in under-provisioned networks. Our numerical evaluation on four different topologies show that exploiting various paths allows to increase the amount of traffic supported by the network without significantly increasing the path length.