
The goal of this project is to construct synthetic workload models to generate realistic ATM cell traffic as input to network simulations. This process involves data collection of network traffic measurements from operational networks (e.g., Ethernet, Internet, ATM), analysis of this data, workload characterization, construction of traffic models, and validation of traffic models against the network measurements. This research is led primarily by Carey Williamson.
To date, we have built a total of four non-trivial network traffic models: a model of MPEG and JPEG compressed video traffic, a model of self-similar aggregate Ethernet LAN traffic, a model of client-side World Wide Web behaviour, and a detailed model of TCP over ATM. These models are used in ongoing simulation projects on ways to improve the performance of TCP over ATM, on congestion control approaches for ATM ABR traffic, and on the statistical multiplexing of self-similar network traffic.
Through collaboration with TRLabs in Winnipeg, we have also been able to collect and analyze ATM network measurement datafrom our local ATM LAN's in Saskatoon, as well as from the CANARIE National Test Network (NTN). To date, we have measurements of: FTP/TCP performance over a wide-area ATM network; JPEG video traffic for sports video in an ATM LAN; MPEG video traffic for entertainment video on a MAC with ATM; measurements of MBONE traffic over the NTN; and benchmark measurements of throughput and delay for several ATM switch vendors.
Comp
Sci. Dept |
University |