| Switching technology continues to be one of the bottlenecks in the development of broadband networks. Virtual Output Queuing (VOQ), where each input maintains one queue for each output, is widely used because it does not have a throughput limitation and does not require a high internal speedup. Scheduling algorithms are used to determine which VOQs will be served in each time slot.; Dual Round Robin Matching (DRRM) Algorithm is a matching scheme with low complexity. We prove that DRRM achieves 100% throughput under uniform i.i.d. traffic. Using a similar approach, the iSLIP algorithm is proved to achieve 100% throughput.; In this thesis, a set of matching algorithms with low complexity and good performance are presented. Most of previous scheduling algorithms try to find as many input-output matches as possible in each time slot, while Exhaustive Service Matching achieves efficiency by minimizing the matching overhead over time. The switching performance, especially under bursty and nonuniform traffic, is significantly improved. Two variations to Exhaustive Service Matching, Limited Service Matching and Exhaustive Service Matching with Hamiltonian Walk (EMHW), are developed to avoid the unfairness under extremely unbalanced traffic. We prove that EMHW is stable under all admissible traffic. The performance of Exhaustive service iSLIP and Exhaustive service DRRM under uniform traffic are analyzed by using an exhaustive service random polling system model.; In optical switches, the reconfiguration overhead is much larger than in electronic switches. We developed two classes of matching algorithms, fixed-size synchronous frame matching and variable-size asynchronous frame matching, so that connections can be updated less frequently to reduce the bandwidth lost.; In order to maximize the network resources, Multi-Protocol Label Switching Traffic Engineering (MPLS TE) is used in IP networks so that traffic can be routed on a path that is not chosen by a standard routing method. In this thesis, a scheme using Traffic Engineering (TE) queues for configured MPLS TE tunnels in every router the tunnel traverses is presented. The performance is simulated and analyzed with a Generalized Processor Sharing system. According to our results, using TE queue leads to lower overflow probabilities for TE tunnel traffic. |
