Qos-based design of high-speed networks

WDM technology can be used to construct high-speed, local-area networks. A new network architecture for creating a WDM client-server network is proposed. It is a novel architecture for asymmetric high-bandwidth applications. It consists of multiple data channels downstream and a single upstream control channel. Detailed point-to-point, connection-setup procedure and internetworking scheme are proposed. The system's performance is analyzed in terms of whether the control channel or the data channels are the bottleneck. The analyses of the request delay, the request's channel-holding time, and the system throughput are also presented. A multi-protocol-label-switching (MPLS)-control-based, fully distributed algorithm to solve the protection problem in IP-over-WDM network is studied. The proposed on-line algorithm includes intelligent and fully automatic procedures to set up, take down, activate, restore, and manage a backup lightpath. By allowing the sharing of backup lightpaths, the algorithm decreases the resource requirement for a backup lightpath reserved for protection while strictly satisfy the connection QoS. The illustrative studies that compare the performance of 1:1, unlimited sharing, and proposed Qos-based sharing backup indicate that the QoS based shared-protection achieves the comparable performance as unlimited sharing, which is much better than the 1:1 back scheme. Modern wide-area networks (WANs), such as the Internet, are heterogeneous. Layered-video is to address the heterogeneity and can distribute video-data to user via multiple-layer multicasting. The problem of computing such data-distribution-paths is NP-complete. We present a new heuristic algorithm called LVMSR with O(Rn2) time complexity and O(R2) message complexity. Simulation results indicate that our algorithm can construct a multicast path with small cost and satisfy all constraints for a randomly generated network topology. In providing video services over a network, the server generally has a finite number of streams (or channels), hence, batching of requests from clients and multicast-technology are widely used. The server must decide when and how to allocate an available channel. A new batching policy and channel-assignment scheme called Token-Tray/Weighted Queuing-Time is proposed. The simulation results show that, compared with other conventional schemes, this scheme achieves the highest revenue and lowest overall loss rate, and the loss rate across movies is fairly uniform and the user delay is reasonably low even at high arrival rate.