Design and analysis of wavelength division multiple access lightwave packet networks

With the advance of fiber optics technology, it is conceivable that we could build multiple access networks on the order of 30 terabits per second (1 terabit = 1,000 gigabits) by using the low-loss band of the optical fiber spectrum (1200-1600 nm). An obstacle is the severe bottleneck at the speed of the electronic interface, typically at a few Gb/s. Wavelength Division Multiple Access (WDMA) eliminates this bottleneck by operating on multiple channels at different wavelengths, with each channel running at full electronic speed.; We first present a mathematical model which closely approximates WDMA multiuser networks with general receiver/transmitter configurations and arbitrary traffic patterns. We first study the case of a uniform traffic matrix and observe that only a small number of tunable transmitters and receivers per station is needed to produce performance close to the upper bound. We then construct a general traffic model and propose an iterative solution procedure. A case of hot-spot traffic is also studied using this model. We find that adding more resources to the hot-spot node helps improve its performance, but only to a limited extent determined by the traffic imbalance.; We next propose a multiple access protocol for a system with a passive star coupler. A station must reserve a wavelength first, then transmit the data on that wavelength. The performance of the protocol is modeled and analyzed for both the infinite and finite population cases. Numerical results show that low delay and high throughput (larger than the electronic speed of a single station) can be achieved. The analysis also shows that the best performance is obtained when the capacities of the reservation channels and the data channels are balanced.; Considering the limitations of star networks in supporting a large number of users in a large area, we then propose a WDMA protocol for a dual bus network. The proposed protocol is a multichannel generalization of the DQDB protocol. An approximate queueing model is built to analyze the system performance. Numerical experiments show that more wavelengths produce higher throughput, but lower efficiency per wavelength. Also we observe that the protocol achieves better fairness than single channel DQDB.; In the end we point out some directions for future research, and conclude that WDMA is a very promising approach to better exploit the enormous bandwidth of optical fiber.