general information
research groups
radioscience seminars
stanford courses
oral defense abstract
industrial affiliates
Oral Defense Abstracts

Date: Monday, December 08, 1997
Time: 3:15 pm (Refreshments at 3:00 pm)
Location: Durand 450

Special University Ph.D. Oral Examination
Characterization Modeling and Multiplexing of Real-Time Video Traffic Over Broadband Networks
Sanjay K. Agrawal
Department of Electrical Engineering

Broadcast-quality MPEG-II digital video is gaining wide-spread use in switched telephony, cable and satellite broadcast domains. Broadband networks, such as Asynchronous Transfer Mode (ATM) and STARNET have gained much attention because of their effectiveness in reserving network resources for the delivery of traffic with different levels of Quality of Service (QoS). Providing a sustained QoS requires reservation of network resources, but reservation of resources typically leads to under-utilization of the network. Dynamic allocation of network resources based on traffic characterization and modeling is necessary to optimize network utilization while providing the desired QoS.

Various techniques have evolved that model and characterize the video traffic for the purpose of estimating fluctuations during steady-state. However, existing approaches fail to model transient characteristics of the video traffic accurately, which are of importance in predicting the consumption of network resources.

In this talk we present a novel technique for modeling video traffic called the Network Queuing Model (NQM). The NQM focuses on modeling the queuing behavior of the traffic rather than simply imitating the fluctuations of source traffic traces. The NQM accurately captures both steady-state and transient behavior of video traffic in network queues enabling, in turn, accurate prediction of required network resources. Our NQM characterization method models traffic as a superposition of different traffic burst components for a more specific description of video during network connection QoS negotiation. We will present NQM performance results and compare them with existing video traffic modeling techniques.

< back to top