In recent years dynamic programming of networks has gained popularity, this is known as software defined networks. The communications protocol OpenFlow has become one of the most important amongst the software defined networks. Nowadays most of the OpenFlow configurations are made by applications, and these applications are manually created by manipulating rules that define network behavior. This can be an error prone process, causing unwanted interactions between the rules. Bifulco and Scheider (2013) proposed a formal definition for the interactions between rules and a detection algorithm, but stated that the algorithm was suited only for OpenFlow applications with just a few hundred of rules or during development. This work presents an improvement of the interactions detection algorithm performance, using well-known data structures to reduce the amount of required operations and a lazy comparison between rules (analogous to lazy initialization). The experiments shown the achieved improvement depends on the composition of the rule set, but the overall improvement was of 41%, and it was determined that the changes in the algorithm represent an alternative approach that suggests the utilization of the algorithm inside an OpenFlow switch and not just for small applications or during development as originally conceived.

Proyecto de Graduación (Maestría en Computación con énfasis en Ciencias de la Computación) Instituto Tecnológico de Costa Rica, Escuela de Ingeniería en Computación, 2015.

Instituto Tecnológico de Costa Rica, Escuela de Ingeniería en Computación