Dynamic scheduling in TWDM-PONs using game theory

• Published in: Optical switching and networking Vol. 33; pp. 103 - 113
• Main Authors: Dalamagkas, Christos, Sarigiannidis, Panagiotis, Kapetanakis, Stelios, Moscholios, Ioannis
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2019
• Subjects: Game theory; Next-generation passive optical network 2; Scheduling; Time and wavelength division multiplexing; Scheduling; Next-generation passive optical network 2; Game theory; Time and wavelength division multiplexing
• ISSN: 1573-4277; 1872-9770
• DOI: 10.1016/j.osn.2017.12.004

Efficient utilization of Passive Optical Networks (PONs) within a broad region of high user numbers, heavy bandwidth demand and large-scale central servicing stations seems a challenging task in explosive high-bandwidth environments. New Generation PONs (NG-PONs) enable multiple channels in both directions and they seem promising towards meeting advanced user applications and services. However, sophisticated bandwidth distribution should be in place to ensure appropriate allocation and efficient management of heavy user demands subject to the available channels. Most of the literature on dynamic bandwidth distribution in NG-PONs usually pertains to heuristic techniques that may be susceptible to expert bias, require substantial knowledge and experience to be applied and their evaluation may focus primarily on minor issues vs. few fundamental major ones. To surpass these limitations game theory can be applied to address efficiently the above heuristic disadvantages and solve the heavy user load-heavy demand over large bandwidth problem in a stochastic way. This work presents the next step forwards using multi-channel multi-stochastic reasoning by utilizing the full channel range to satisfy fairly user requests and high network traffic within a NG-PON2 network. We present the full algorithm as applied to standardized NG-PON2 as well as a series of evaluation experiments to prove its applicability in terms of fairness, goodput and delay. According to the simulation results, the proposed scheme succeeds to provide a fair bandwidth distribution, when triggered, while it is able to a) reduce the experienced delay from 1 to 6 msec and b) double the offered goodput compared to the pure allocation scheme.