Estimating available bandwidth for real-time supermedia applications

The emerging development of remote sensing and control technology gives rise to new types of real-time networking applications that require "pro-active" quality of service guarantees. This dissertation specifies Quality of Service parameters characterizing requirements for successful execution of a given real-time tele-operation over the Internet. The dissertation also provides a statistical model that outputs the likelihood of a failure of remote real-time task execution based on the history of behavior of an end-to-end network path and the task requirements. We provided and compared both parametric and non-parametric approaches, conducted their experimental evaluation and also introduced some mechanisms for improvement of present measurement technologies of an end-to-end quality of a network path.;Available bandwidth (spare capacity) of a network path is a key characteristic of the network path quality. We regard available bandwidth as a random time-series with an existing auto-correlation structure that is realistic and makes channel quality estimation accurate on real paths over the Internet.;We also consider a set of pro-active measures to save remote real-time task execution from a failure once the likelihood of such a failure exceeds an acceptable threshold. In particular, we give a solution to the Reroute Sequence Planning problem: to choose an optimal sequence of data-streams reallocations across different network paths once the QoS of old paths are unacceptable. Optimal reallocation minimizes extra capacity of a network link that is required for carrying out the reallocation procedure.