Institute for Systems Research

We examine the question of how to model HSTN traffic for network dimensioning and performance prediction, and in particular, how far ahead into the future a traffic model can be expected to accurately function. We investigate these issues by directly comparing four of the most likely candidate statistical distributionshe exponential, log-normal, Weibull and Pareto. These distributions are fit to two key traffic parameters from real HSTN traffic traces (connection interarrival times and downloaded bytes), and their relative fits are compared using statistical techniques. We further compare traffic models built using these distributions in a simulated environment; comparing performance predictions (over a number of metrics) obtained from these models to the actual results from our real-world traffic traces.

In the first part of this thesis, we analyze theimplementation feasibility and overhead of these schemes to determinewhether this overhead represents an obstacle to deployment. To this end,we employ a novel approach with real traffic traces from the Internet anda passive gateway simulator.

Having established the feasibility of suchalgorithms, we turn our attention to buffer management techniques in thepresence of fair resource allocation. Our specific focus is on developingan effective buffer management technique for satellite networks. Thelimitations of existing schemes lead us to propose a new buffer managementscheme designed with our problem space in mind.

Finally, we look at aclass of satellite enhanced gateways, termed as connectionsplitting or spoofing gateways, proposed for implementing highperformance satellite systems. The specific design issues peculiar tothis class of gateways are analyzed. A novel architecture for fair resourceallocation in spoofing gateways is then proposed. The efficacy of ourarchitecture in providing fairness and protecting adaptive flows againstmisbehaved and non-adaptive flows is demonstrated by means of simulation.

This is a major departure from all other broadcast optimization schemes which are handicapped by their total reliance on complete knowledge of both ``hits'' and ``misses''. We also show that the proposed adaptive scheme performs effectively even under very dynamic and rapidly changing workloads. Extensive simulation results demonstrate both the scalability and versatility of, the technique. Another basic result obtained in this paper is that the overall, system's throughput depends only on the size of the hot-spot and not on the volume of the workload. This has far reaching implications for very large scale and high volume wide area information systems.

Optimal dynamic routing in such mixed-media (voice/data-integrated) networks under Markov Queueing Modeling has been developed and solved. Routing problems in such a domain usually lead to a weighted-sum minimization or a minimax problem. A new approach to obtain the trade-off curve of multiple-objective optimization is outlined. With a numerical optimization package, we can plot the trade-off curve exactly.

Both a centralized and a distributed implementation of this problem with Kalman filter techniques and Equilibrium Programming are presented. These techniques allow the control of a large and stochastic network by communicating with a group of communication managers in a parallel manner.

For ATM- based hybrid networks, an economic model with the objective of maximizing the ﲳocial welfare has been adapted. This model can be developed into a form of a two-player game. With a little modification and adaptation, we will be able to solve the joint problem of access control and routing in both the weighted-sum formulation and the economic formulation.