Dynamic reallocation of SLA parameters in passive optical network based on clustering analysis

The introduction of new services as well as the growth in the number of communication terminals in the last years has led to an exponential growth of data traffic in both fixed and mobile networks. Passive Optical Networks (PONs) offer high bandwidth services to service providers customers. However, due to the dynamicity of users traffic patterns, PONs need to rely on an efficient upstream bandwidth allocation mechanism to define for each customer the amount of data that needs to be transmitted at a specific time. This mechanism is currently limited by the static nature of Service Level Agreement (SLA) parameters which can lead to an unoptimized bandwidth allocation in the network. In this paper, we propose a novel mechanism for optimizing the allocation of upstream Gigabit-capable Passive Optical Networks (GPON) resources based on the dynamic adjustment of some SLA parameters according to customer's estimated traffic patterns. Clustering analysis is used to differentiate customers according to their bandwidth utilization based on real-time and historical data. Three user classes are taken into account: heavy, light and flexible. Our work considers two fundamental clustering algorithms, namely K-means, a very well-known partitioning method and DBSCAN, one of the most common density-based clustering algorithms. An experimental study is conducted to evaluate the two algorithms and select which one can be the most suitable for the differentiation of user classes.