The max–min fair approach on dynamic bandwidth allocation for XG‐PONs

Abstract

XG-PON is the latest ITU-T PON standard for passive optical networks (PONs), which provide broadband access to the end user at line rates of up to 10 Gb/s. The most important issue that is not strictly defined in the standard and is left to vendors to implement is the allocation of the shared upstream bandwidth. An efficient allocation mechanism must be applied on the upstream direction in order to assure acceptable packet latency and fairness among multiple subscribers. Within this context, we propose an XG-PON-oriented dynamic bandwidth allocation algorithm that implements a maxmin fair scheme to distribute the available upstream capacity to all the connected optical network units (ONUs) at the PON. Because frequent bandwidth granting and queue occupancy reporting account for low latency, our algorithm adopts a method in which every ONU is granted a portion of bandwidth at every downstream frame. The bandwidth grant is either calculated directly from the latest occupancy report or inferred from previous reports by means of traffic prediction. Our simulation results show that the proposed algorithm achieves fair bandwidth distribution among the ONUs along with low packet latency and latency jitter, irrespectively of the ONUs's spatial separation from the optical line terminal. Copyright

Burst‐by‐burst dynamic bandwidth allocation for XG‐PONs

“…XG-PONs, which are standardised by ITU [15], have not received equally significant attention and only a few DBA algorithms have been proposed exclusively for XG-PONs [16][17][18][19][20][21]. XG-PONs are engineered for the transport of time-critical data and the ITU-T recommendation enforces stringent timing requirements on their operation [22], thus inherently limiting the data latency.…”

“…Despite the fact that the existing DBAs provide a pre-determined degree of SLA, their main drawback is that they are not closely adapting to the instantaneous traffic demands of ONUs and bandwidth waste can be observed, especially for best-effort traffic. To address this issue, we recently proposed a max-min fair DBA that is suitable for XG-PONs and distributes resources among ONUs in an increasing order of demand [21]. Our previous work, however, did not take into account the start and finish times (FT) of ONU bursts, and if an ONU was partially served then its remaining request had to compete for bandwidth with newly arrived requests in the upcoming frame.…”