By offloading the control plane to powerful computing platforms running on commodity hardware, SDN unleashes the potential to operate computation intensive machine learning tools and solve complex optimization problems. This paper studies such an opportunity under the framework of the centralized SDN Admission Control problem. To our knowledge, online algorithms for this admission control problem, have been seldom applied despite the interesting guarantees that they offer. We review and adapt some of the key algorithms from the literature, and evaluate them under different traffic conditions to understand and highlight their strengths and weaknesses. Due to the unpredictable nature of network traffic, we verified that it is impossible to know a priori which algorithm can achieve the best performance under specific scenarios. We thus argue that the computation power of SDN controllers enables the implementation of machine learning techniques which execute all the algorithms in parallel and select sequentially the seemingly best one. We then review and evaluate the performance of some expert meta-algorithms, which turn out to further improve the admission control performance.
Mots clés :