Study On The Assembly Algorithms In Optical Flow Switching Networks

Optical Flow Switching (OFS) technology plays an important role as one active research area in all optical networking technologies. As an efficient way of providing high-capacity, high-bandwidth and real-time transmission, OFS networks can set up fast connections for dynamic requests flexibly, and perform well with unique characteristics and advantages. Flow assembly is an important component of the source node in OFS networks that it assembles the input traffic from the higher layer into buffers according to the same attribute. The flow assembly algorithm largely affects the network performance, and has important practical implications.The assembly algorithms are the research objective in this thesis, and OMNeT++is the simulation tool. Firstly, according to the design principle and hierarchy of OFS networks, two methods to set up the lightpaths are studied based on the centralized transmission mechanism and the distributed transmission mechanism, named entropy-assisted fast scheduling algorithm and multi-destinations control protocol (MDCP) scheduling algorithm. MDCP algorithm becomes a foundation for the experimental platform, and contributes to the simulation of the assembly algorithm. Based on the above knowledge, some assembly algorithms are studied including the time-based, threshold-based, hybrid and adaptive algorithms. Moreover, there is a detailed introduction to the basic idea and the unique characteristics of each assembly algorithm, which gives an exposition of the parameter settings.Both National Science Foundation NETwork (NSFNET) and full-mesh networks topology are adopted for simulation using OMNeT++. Some theories are verified by the simulation, including the influence of queueing delay and throughput due to different assembly algorithms, and also the flow size, flow generation period, and setup period due to the load and threshold. The simulation results prove the algorithm based on hybrid threshold combines the advantages of the two algorithms based on fixed threshold, while adaptive algorithm performs better.