Dynamic Bandwidth Allocation Algorithms In Long Reach Passive Optical Networks

Long Reach Passive Optical Network (LR-PON) can simplify the network structure and provide broadband acess services with low cost for more users, which is thought to be one of the most promising candidates for next generation passive optical network (NG-PON). Besides, LR-PON can extend the coverage of Ethernet Passive Optical Network (EPON) from 20 km to 100 km or more, which results in an increased propagation delay. To make better use of upstream channel of LR-PON and also to reduce the packet delay of upstream packets, the design of an effective Dynamic Bandwidth Allocation (DBA) algorithm has become the urgent problem required to be solved in LR-PON.Firstly, we introduce the history of development, the network structure and the research status of LR-PON. And then, the major DBA algorithms in LR-PON and their advantages and disadvantages are generally analyzed. Additionally, with a consideration of heterogeneous propagation in LR-PON, some modifications are made on Multi-Thread Polling (MTP) algorithm in order to reduce the void time in upstream channel, and an Improved Multi-Thread Polling (IMTP) algorithm is proposed in this paper. In IMTP, all ONUs are divided into two groups, namely short distance ONUs and long distance ONUs, based on the distance between them and the OLT. The main idea of IMTP is to allocate the upstream void time between the time for the short distance ONU to send the last packet and the time that the OLT receives the first packet of long distance ONU. This part of void time is allocated between short-distance ONUs according to their reported bandwidths. By doing so, the upstream void time can be reduced and the channel utilization as well as the packet delay performance can be improved. Moreover, considering the limitation that MTP algorithm can not adjust the thread number in the polling cycle according to the network load dynamically, we propose an Adaptive Multi-Thread Polling (AMTP) algorithm. In AMTP, OLT changes the thread number used according to the upstream packet delay. In the situation of low network load, if the upstream packet delay is too big, AMTP increases the polling thread number, otherwise, it decreases the thread number.On the contrary, when the network load is high, in case of big packet delay, AMTP reduces the polling thread number, and vice versa. In this way, the upstream packet delay together with the upstream channel utilization performance can be improved. Finally, by setting up the LR-PON simulation platform in OPNET, IMTP and AMTP algorithms are analyzed and compared with the existed DBA algorithms. Simulation results show that, the proposed algorithms can achieve a lower packet delay and delay jitter as well as a higher upstream channel utilization, and is better than than the existing algorithms.