Measurement-based traffic modeling and network resource management

Given the high complexity and large scale of inter-networking today, network development calls for network simulation analysis based on sophisticated traffic generators and careful design of various aspects of network controls which optimize the network resource utilizations. We strive to address various aspects of algorithm design and resource management issues ranging from algorithm design for measurement-based traffic modeling at cell/packet level, link bandwidth sharing for providing differentiated services at call/flow level, to flow classification algorithm design with resource management for flow-based hybrid switching systems at both cell/packet level and call/flow level.; In measurement-based traffic modeling, we develop fast algorithms for a traffic measurement, modeling and queueing analysis tool called Statistical Matching And Queueing (SMAQ) tool. We show that, with the proposed fast algorithms, SMAQ becomes a practically powerful tool, which is readily attractive to field traffic engineers. In providing differentiated services over the Internet, we bring up an issue related to the design of link bandwidth sharing policies and propose a possible solution. This study offers insights on the design of effective link bandwidth sharing policies and pricing schemes. One of the key issues for flow-based hybrid switching systems is the design of effective flow classification algorithm. We propose a new concept of flow classification for resource management and design an effective adaptive flow classification algorithm which optimizes the resource utilizations in a highly dynamic environment. The proposed design principle can be applied to many other network control problems which involve resource allocation/balancing in a dynamic environment. As an industry standard, Multiprotocol over ATM (MPOA) holds the promise to be able to effectively transfer multiprotocol traffic, especially, IP traffic over ATM networks. However, its performance and flow classification design have not been well studied. We formulate a framework to address these issues. Concerns with regard to MPOA performance are clarified and an effective flow classification algorithm in the context of resource management is developed for MPOA.