Dynamic bandwidth allocation in CDMA-based passive optical networks

Abstract

Fiber to the home (FTTH) technology is an attractive solution for providing high bandwidth from the Central Office (CO) to residences and small-and medium-sized businesses. The emergence of Internet Protocol-based communication within households such as VoIP, IPTV, video conferencing, and high definition multimedia shows that there is a need for high-capacity networks that can handle differentiated services. By providing an optical fiber link to a household where the optical network unit (ONU) is located, there will be a tremendous increase in information capacity with respect to Digital Subscriber Line and cable modem technologies that are currently in place. In access networks, Passive Optical Networks (PON) are rapidly replacing copper-based technologies due to a wide range of benefits, one of which is having the capability to transmit data at a higher rate and reach further distances without signal degradation. Under the PON family of technologies, Ethernet PON (EPON) was developed and is specified in the IEEE 802.3 standard outlining the framework that can deliver voice, data, and video over a native Ethernet port to businesses and residential customers. An increasingly important subject to network operators is Quality of Service (QoS). Although the EPON specification provides mechanisms for supporting QoS, it does not specify or define an algorithm for providing QoS. Rather it is up to the CO to design and implement an appropriate algorithm to meet the specifications of services that are offered to their clients. Researchers have extensively studied bandwidth allocation in EPON where the challenge is to develop bandwidth allocation algorithms that can fairly redistribute bandwidth among ONUs based on their demand. These algorithms were developed for the uplink direction, from ONUs to CO, in a network where only a single ONU is permitted to transmit at a time. Another well-established PON technology is Optical Code-Division Multiple Access PON (OCDMA-PON). In recent years, it has become more economical due to hardware advancements and it has gained a lot of attention due to its benefits over EPON. The most attractive benefit of OCDMA-PON is that multiple ONUs may transmit to the CO simultaneously, depending on a number of constraints, whereas EPON is limited to a single ONU transmission at a time. In this thesis, we develop a dynamic bandwidth allocation algorithm called Multi-Class Credit-Based Packet Scheduler (MCBPS) for OCDMA-PON in the uplink direction that supports the Internet Protocol (IP) Differentiated Services and takes advantage of the simultaneous nature of OCDMA. The IP Differentiated Services specifications stipulate the following traffic classifications: Expedited Forwarding for low latency, low packet loss, and low jitter applications; Assured Forwarding for services that require low packet loss; and Best Effort which are not guaranteed any bandwidth commitments. MCBPS incorporates the use of credit pools and the concept of a credit bank system to provide the same services as EPON by assigning ONUs specific timeslots to transmit data and also by specifying the amount of bytes from each class. MCBPS is a central office based algorithm that provides global fairness between Quality of Service (QoS) classes while also ensuring that at any given moment the desired number of simultaneous transmissions is not exceeded. We demonstrate through simulation that MCBPS algorithm is applicable in both EPON and OCDMA-PON environments. An in-house simulation program written in the C programming language is used to evaluate the effectiveness of the proposed algorithm. The MCBPS algorithm was tested alongside a benchmark algorithm called Interleaved Polling with Adaptive Cycle Time (IPACT) algorithm to compare network throughput, average packet delay, maximum packet delay, and packet loss ratio. From the simulation results it was observed that MCBPS algorithm is able to satisfy the QoS requirements and its performance is comparable to IPACT where the simultaneous transmission is limited to one. The simulation results also show that as the number of simultaneous transmissions within the network increases, so does the bandwidth. The MCBPS algorithm is able to re-distribute the scaling bandwidth while ensuring that a single ONU or QoS class does not monopolize all the available bandwidth. In doing so, through simulation results, as the simultaneous transmissions increases, the average packet delay decreases and the packet loss ratio improves.