| Meeting Quality-of-Service (QoS) requirements for various services in the broadband integrated services packet networks has been very challenging to network designers. Broadband integrated service packet networks rely on traffic scheduling algorithms within individual switches or routers to provide a wide range of QoS. The main functions of the traffic scheduling algorithm are (1) scheduling the service and (2) buffer management. Several papers have addressed different scheduling algorithms. One of the main objectives of the scheduling is to achieve fairness in the amount of service provided to the competing sessions, in which the share of service is proportional to some specified weighting factor. In addition, packets from different sessions tend to be interleaved with each other as they are multiplexed onto a transmission link. Consequently, user traffic's burstiness is reduced, which in turn reduces the network congestion probability and increases the network throughput. Generalized Processor Sharing (GPS) is the ideal fluid reference for packet-fair-queuing scheduling disciplines; a class of the packet-by-packet version of scheduling algorithms are proposed, with different tradeoffs between complexity and performance.; In our first work, WFI packet shaper, we generalize the leaky-bucket algorithm to handle variable-length packets. We show that a leaky-bucket size can be determined at different checking points: the beginning of the last departed packet, the end of the last departed packet, or the beginning of the next packet that can start transmission without violating the peak rate requirement. Second, we study the problem of scheduling packets in the non-blocking Combined Input-Output Buffered Switches. We introduce a scheme, called Switch Hierarchical Link Sharing (SHLS), to provide delay bound guarantee for real-time traffic and distribute excess bandwidth fairly according to hierarchical link sharing.; Finally, we propose a simple core-stateless scheme, called the Core-Stateless Shaped Virtual Clock (CSSVC) algorithm, to provide the guaranteed delay bound and minimum WFI without keeping per-flow state information at core nodes. The implementation of both edge routers and core routers in the CSSVC scheme is simplified by using the scheme. The edge router that inserts the state variables for each packet entering the core network does not need to know the path traversed by a flow. (Abstract shortened by UMI.)... |
