Efficient Dynamic Bandwidth Allocation in Long Reach Passive Optical Networks

Abstract

Long Reach Passive Optical Networks (LR-PONs) concept is offered as a cost-efficient solution to bring the fiber to the Home, Office, Curb, Building etc. (FTTX). It extends the reach and capacity of traditional PON from ~20 km to ~100 km using the optical amplifiers. It merges the metro and access domains by reducing the number of central offices (CO). This approach reduces capital and operation expenditure by reducing number of components, along with the network power consumption and design complexity. Both PON and LR-PON use one downstream and one upstream channel. In downstream direction (OLT to ONUs) data is transmitted in broadcast manner and in upstream direction (ONUs to OLT) time division multiple access (TDMA) approach is used to avoid collision [1]. In TDMA approach, timeslots are dynamically assigned to each ONU using multipoint control protocol (MPCP). Transmission-messages of MPCP are exchanged between the optical-line terminals and optical network units before the actual data transmission. By extending the reach, these messages are delayed due to increased Round Trip Time (RTT) which degrades the performance of dynamic bandwidth allocation (DBA). Design of Efficient DBA algorithm is required to overcome this challenge. This thesis presents a survey of Dynamic Bandwidth allocation (DBA) schemes for Long Reach PONs with their pros and cons and also proposes a hybrid scheme of inter and intra ONU scheduling mechanisms for delay reduction in Long reach PONs. Our proposed scheme combines Efficient inter thread scheduling (EIS) and Priority Swapping mechanisms. We used EIS scheme for implementing class of service (CoS) by incorporating Priority Swapping (PS) in it. We categorize the traffic into three classes i.e. Expedited Forwarding (EF), Assured Forwarding (AF) and Best Effort (BE).We have selected Priority Swapping (PS) for intra ONU scheduling. It simply swaps the priority of queues based on the buffer size of these classes. For validation of proposed scheme, self-similar traffic using pareto distribution is generated. Delay is analyzed by varying network load. We compared our proposed scheme with PS, Single thread polling, EIS schemes and have found that it shows better results in terms of delay.