9314387 Li The principal investigator will explore a new system of traffic modeling and network control in high speed networks. The work will build a foundation for integrating traffic measurement and queueing analysis. The investigator will consider three stages of traffic modeling. He will 1) use advanced signal processing techniques in traffic measurement to collect the traffic power spectrum and distribution, 2) develop a programming approach for constructing a single Markov chain to match optimally with the power spectrum and distribution, and 3) evaluate the corresponding queue length and loss rate performance in a finite-buffer overload- controlled system. The investigator will design a software package to accomplish all of the three-stage computation tasks. For the network control, the investigator will use each node in a high speed network to take on-line measurements. On a local link level, such measurements will be used for congestion control, input rate control and simplify greatly the design of network-wide traffic flow control, dynamic routing, and call admission control. On a link level, the investigator will introduce a simple effective method for dynamic link capacity allocation based on the on-line observation of filtered input traffic low in the network. THe key advantages of such a control scheme are the ability to improve the network performance by controlling the low frequency traffic flow and the simplicity of operation by overlooking the queuing process at each node. ***
