Buffer-efficient RTA algorithms in optical TDM networks

Optical Time-Division Multiplexing (O-TDM) requires no optical header processing and could be, in the near future, a more practical alternative to optical packet-based switching schemes such as Optical Burst Switching and Optical Packet Switching for providing a finer bandwidth granularity than that can be provided by pure Wavelength Division Multiplexing (WDM) networks. As mature technologies for optical buffering are not yet available, we believe that in an O-TDM network, an important design criterion will be the minimization of the buffer requirement. The theme of this thesis is the exploration of Routing and Time-Slot Assignment (RTA) strategies and network architectures for O-TDM networks with the purpose of reducing the buffer requirement without sacrificing network capacity. In the first part of this thesis, we first formulate a Markovian model for the arrival and departure process of O-TDM circuits in a single node. We then establish an analytic procedure for the computation of 'bandwidth blocking' and 'buffer blocking', and compare the relative contributions of the two types of blocking to the overall blocking probability. Here we show that we can set a limit to the maximum buffering delay that is allowed in a single node without adversely affecting the overall blocking probability. Then we go beyond a single node and explore RTA algorithms for O-TDM mesh networks. An Expanded Shortest Path (ESP) algorithm is devised such that by expanding each network node into an expanded set of mini-nodes, we can flexibly solve the RTA problem under various bandwidth and buffer cost assumptions. Since the result for the single node case shows that we can reduce the buffer holding time without degrading the overall call blocking probability, we formulate also a trimmed ESP algorithm with greatly reduced complexity. In the second part of this thesis, we shift our focus to O-TDM ring networks without Time Slot Interchangers (TSIs). We show that the intrinsic propagation delay in a TSI-free O-TDM ring network will lead to formation of 'slot cycles' which link together the time-slots in the ring. By studying the concept of 'slot cycle' and the algebraic relationship between the frame size and the total propagation delay of the ring, we discover an O-TDM ring architecture where all slots belong to one slot cycle. We show that by using carefully designed RTA strategies, our proposed architecture can achieve many of the widely used capacity bounds that have been established for WDM rings with wavelength converters. An example is the ⎡N / 4⎤bound for the N-node single-port bidirectional ring. In the third part of the thesis, we extend our results in the second part to the case of Single-Connected Double Ring. We believe that interconnected ring may be needed for extended geographical coverage in an O-TDM ring network. Especially, we explore the associated slot cycle properties for the Single-Connected Double Ring. We show that some important properties of an isolated TSI-free ring still hold in an interconnected ring, but other properties will change. We analyze the capacity required to set up a traffic matrix for a given load and show its corresponding time slot assignment algorithm. We also present a generalized Tucker's algorithm that can be used for cases where TSIs are deployed to reduce the capacity requirement. Results in this thesis will enable system designers to make the best tradeoff between bandwidth efficiency and buffer efficiency in an O-TDM network