| High-speed ATM networks are expected to support various types of applications, each with their own Quality of Service (QoS) requirements. Each application may generate one or more classes of traffic such as voice, video, still images, and data. The problem of utilizing the network efficiently and not (statistically) violating the negotiated QoS requirements of existing applications has been the subject of extensive research. In this thesis we study aspects of this problem that involve scheduling and flow control.;We characterize the different traffic classes into two main types: real-time and controlled traffic. The real-time traffic consists of applications with stringent delay and loss constraints such as video and voice traffic. The controlled traffic, composed mainly of current Internet data traffic, can tolerate large delays but requires lossless transmission. We propose an effective hop-by-hop feedback flow control framework to support these types of traffic classes and maintain high network utilization. The proposed flow control mechanism provides lossless service for the controlled traffic. To guarantee QoS requirements for the real-time traffic, we develop optimal scheduling algorithms based on different delay and priority constraints of the traffic. We define and investigate the complexity of real-time scheduling algorithms, then develop simpler heuristic scheduling algorithms based on our optimal algorithms. We use an analytical model to estimate the loss rate performance and gain insight on the need for scheduling and when it is most beneficial.;Finally, we present some concluding remarks on how this research might be extended in different directions. |
