Abstract:

Ethernet passive optical networks (EPONs) address the first mile of the communication infrastructure between the service provider central offices and the customer sites. As a low-cost, high-speed technology, EPONs are deemed as the solution to the bottleneck problem of the broadband access network. A major feature of EPONs is the utility of a shared upstream channel among the end users. Only a single optical network unit (ONU) may transmit during a timeslot to avoid data collisions. In order to provide diverse quality of service (QoS), the bandwidth management of the upstream channel is essential for the successful implementation of EPONs, and thus, an efficient medium access control is required to facilitate statistical multiplexing among local traffics.

This dissertation addresses the upstream bandwidth allocation over EPONs. An efficient mechanism, i.e., limited sharing with traffic prediction (LSTP), has been proposed to arbitrate the upstream bandwidth among ONUs. The MultiPoint Control Protocol (MPCP) messages, which are stipulated by the IEEE 802.3ah Ethernet in the First Mile (EFM) Task Force, are adopted by LSTP to facilitate the dynamic bandwidth negotiation between an OLT and its associated ONUs. An ONU reports the already enqueued data, and the OLT conducts on-line traffic prediction to forecast the incoming data during the waiting time. The bandwidth assignment decision is arbitrated based on the bandwidth requirement of the agreed service contract.

With respect to the performance evaluation, theoretical analysis on data loss, data delay, and queue length has been conducted. The quantitative results demonstrate that 1) the innovative LSTP mechanism dynamically allocates the upstream bandwidth among multiple ONUs; 2) the traffic predictor delivers satisfactory prediction for the bursty self-similar traffic, and thereby, contributing to the reduction of data loss, data delay, and queue length; and 3) the bandwidth arbitration at the OLT effectively restricts the aggressive bandwidth competition among ONUs by adopting the service level agreement (SLA) parameter as the upper bound. Aside from analysis, the LSTP mechanism has been substantiated by experimental simulations.

In order to differentiate the service provisioning among diverse users, LSTP is further enhanced with the support of dynamic bandwidth negotiation based on multiple queues. The incoming traffics are first classified into three classes, and then enqueued into the corresponding queues. A traffic predictor is dedicated to one class of traffic. Service differentiation among classes is provided by the combination of queuing and scheduling at the ONU side. At the OLT side, the bandwidth allocation for each class of traffic is based on the reported bandwidth requirement and is upper-bounded by the SLA parameter. Experimental simulations have justified the feasibility of providing service differentiation over broadband EPONs.

Committee Members:

Dr. Nirwan Ansari, Advisor, Professor, NJIT

Dr. Swades De, Assistant Professor, NJIT

Dr. Kevin. W. Lu, Chief Scientist, Telcordia Technologies, Inc.

Dr. Roberto Rojas-Cessa, Assistant Professor, NJIT

Dr. Sirin Tekinay, Associate Professor, NJIT