Using Genetic Algorithms to optimise dynamic power saving in communication links subject to quality of service requirements

•This paper is a Genetic Algorithm (GA) and our Slowing Mechanism (SM) integration to provide insight of how a GA optimisation can be employed in a network environment to optimise parameters according to a set of samples that were collected in real-time.•The research demonstrate that real-time optimisation successfully performed by the GA and the SM saves a considerable amount of power by considering hardware, power, variable load and performance aspects of the networking.•The paper offers a packet delay threshold for the applications in consideration of the non-delay tolerant VoIP. According to the application's maximum allowed delay the maximum per hop packet delay should not exceed 7.5 ms.•Administrators can set any number for the threshold and without any modification; the SM is able to provide a near optimal solution dynamically respond to the variable Incoming Traffic pattern using the GA optimisation for the new delay threshold.•A 61% power saving was achieved by a 4.5 ms 97.5th percentile packet delay without packet loss for the 30% utilised typical metro-core link with the given typical stable traffic and default parameters. This is reduced down to 17.5% saving by 6.3 ms 97.5th percentile packet delay and less than 0.001% packet loss for the 36% utilised link with the bursty traffic and default parameters.

Network devices, meeting increasing workload demand, are not efficiently Power-Workload Proportionate and consume a considerable amount of power even when the workload (utilisation) is low. This work proposes a novel Slowing Mechanism (SM) that provides Power Workload Proportionality for a wired network equipment to reduce power consumption. The Slowing will be achieved by adjusting the Operational Rate (OPR) of components according to traffic load. To meet applications’ (VoIP, Data and Video) performance requirements, a Safety Gap (SG) is proposed in the Slowing Mechanism. Many parameters need to be carefully set for performance requirements within Slowing Mechanism. A Genetic Algorithm (GA) optimisation dynamically set to respond to the variable incoming traffic pattern determines these parameters. Thus, this work is a GA and the Slowing Mechanism integration to provide an insight into how GA optimisation can be employed in a network environment, and to optimise parameters in real-time.The results demonstrate that a considerable amount of saving is achievable. With the default hardware configuration, the SM optimises the parameters and offers a saving of over 60% for typical stable traffic, with acceptable packet delay and no packet loss. This saving is reduced to 17% saving for a bursty traffic pattern with acceptable performance degradation.