Least constraining time-slot allocation in GMPLS optical TDM networks and optimization of optical buffering.

摘要

Optical Time Division Multiplexing (OTDM) in optical networks is a bandwidth sharing technique that organizes access to a shared wavelength in equal time-slots organized in repeated frames. In this case, a transmission channel can be established at the time-slot level instead of using the full wavelength. The main advantage of this technique is to allow several low speed communication channels to coexist on the same high speed optical wavelength, and hence to make effective use of the enormous bandwidth available on a single wavelength. On the other hand, the major problem with OTDM is the time slot continuity constraint in an OTDM channel, which is similar to the wavelength continuity in a WDM channel. Due to this constraint, time slot contentions can exist in the network if proper scheduling and slot reservation techniques are not employed. Basically, the adopted time slots must be free on all links throughout the communication route in order to successfully reserve a communication channel. In addition, to mitigate the effect of slot continuity constraint on bandwidth utilization, appropriate time slot buffering (or interchanging) is often employed. Previous work assumed the deployment of Optical Time-Slot Interchangers (OTSI) to solve the contention problems regardless of their industrial feasibility. In addition, other work considered very basic reservation schemes to achieve proper scheduling, such as the First Fit (FF), Random Fit (RF), and Least Loaded (LL) schemes. In this thesis, we propose a new time-slot reservation scheme for OTDM networks without buffering to significantly improve the performance and eliminate the buffering overhead. It is the Least Constraining (LC) slot reservation scheme which allocates resources having the lowest possible constraints on other resources in the network. In addition, we define a distributed scheme to deploy the LC approach in GMPLS networks, and prove that the same performance level can be maintained by a distributed signaling protocol. Finally, we propose an optimized optical buffering technique to achieve close to optimum performance when the LC reservation approach is not used. It helps in building effective time slot synchronization devices uses between adjacent node pairs.