| This thesis describes a framework for bandwidth allocation and scheduling in the Agile All-Photonic Network (AAPN). This framework is also applicable to any single-hop communication network with significant signalling delay (such as satellite-TDMA systems). Slot-by-slot scheduling approaches do not provide adequate performance for wide-area networks, so we focus on frame-based scheduling. We propose three novel fixed-length frame scheduling algorithms (Minimum Cost Search, Fair Matching and Minimum Rejection) and a feedback control system for stabilization.; MCS is a greedy algorithm, which allocates time-slots sequentially using a cost function. This function is defined such that the time-slots with higher blocking probability are assigned first. MCS does not guarantee 100% throughput, thought it has a low blocking percentage. Our optimum scheduling approach is based on modifying the demand matrix such that the network resources are fully utilized, while the requests are optimally served. The Fair Matching Algorithm (FMA) uses the weighted max-min fairness criterion to achieve a fair share of resources amongst the connections in the network. When rejection is inevitable, FMA selects rejections such that the maximum percentage rejection experienced in the network is minimized. In another approach we formulate the rejection task as an optimization problem and propose the Minimum Rejection Algorithm (MRA), which minimizes total rejection. The minimum rejection problem is a special case of maximum flow problem. Due to the complexity of the algorithms that solve the max-flow problem we propose a heuristic algorithm with lower complexity.; Scheduling in wide-area networks must be based on predictions of traffic demand and the resultant errors can lead to instability and unfairness. We design a feedback control system based on Smith's principle, which removes the destabilizing delays from the feedback loop by using a "loop cancelation" technique. The feedback control system we propose reduces the effect of prediction errors, increasing the speed of the response to sudden changes in traffic arrival rates and improving the fairness in the network through equalization of queue-lengths. |
