A new prediction based technique for Quality of Service aware Traffic Engineering

• Published in: 2019 27th Telecommunications Forum (TELFOR) pp. 1 - 4
• Main Authors: Tomovic, S., Radusinovic, I.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-11-2019
• Subjects: prediction; reconfiguration overhead; SDN; traffic engineering
• DOI: 10.1109/TELFOR48224.2019.8971029

In this paper, we propose a new technique for Quality of Service (QoS) aware traffic engineering (TE) that efficiently handles traffic variations with a small reconfiguration overhead. In particular, we consider the problem of minimizing Maximum Link Utilization (MLU) in Software Defined Networks (SDNs) by periodically reconfiguring load balancing weights for pre-installed paths. The optimal load balancing scheme is computed for each traffic class, by taking into account specific QoS constraints and a preferable time-average reconfiguration rate from the perspective of the network operator. In order to avoid frequent reconfigurations, a traffic prediction model is used to forecast a traffic matrix over a predefined time-horizon. In addition, the Lyapunov drift plus penalty algorithm is applied at each optimization instant to find the most efficient trade-off between the network MLU and a drift function that models the time-average reconfiguration rate constraint. The efficiency of the proposed solution has been evaluated via simulations on Sprint network topology. The obtained results prove that the proposed technique is able to efficiently handle traffic variations with a small number of route changes, even with a simple traffic prediction model.