Optimised QoS-Aware DBA Mechanisms in XG-PON for Upstream Traffic in LTE Backhaul

Abstract

Passive Optical Networks (PON) are viewed as an attractive choice to provide flexible and cost-efficient backhaul for Long Term Evolution (LTE) cellular base stations (eNB). PONs, such as the 10-Gigabit capable PON (XG-PON), use a Dynamic Bandwidth Allocation (DBA) mechanism to multiplex the shared upstream medium between competing sending nodes. Due to the complex definitions of QoS for DBA in XG-PON, it is a challenge for the XG-PON to respect the Quality of Service (QoS) requirements for different aggregated upstream applications in the LTE backhaul, in particular voice, live video and best-effort Internet traffic. In this paper, we first evaluate two recent XG-PON-standard-compliant DBAs - XGIANT and Efficient Bandwidth Utilisation (EBU) - for mean queuing-delay performance, with regard to priority and fairness, when the realistically-generated upstream voice, video and best-effort applications are aggregated at the evolved Node B (eNB) in LTE. We show that neither XGIANT nor EBU satisfy the priority and fairness requirements for mean queuing-delay. We propose and evaluate two optimised DBAs - Deficit XGIANT (XGIANT-D) and Proportional XGIANT (XGIANT-P). Our evaluations of the optimised DBAs, in the ns-3 network simulator, show that both XGIANT-D and XGIANT-P are able to ensure strictly prioritised, low and fair mean-queuing-delays for eNB-aggregated voice, video and best-effort traffic in two loaded conditions in XG-PON upstream. XGIANT-D and XGIANT-P, when compared with XGIANT and EBU, also ensure lower packet losses for eNB-aggregated upstream traffic in the XG-PON-based LTE backhaul.