Firass Abi-Nassif

Offered Load Estimation for a Multimedia Cable Network System

Friday, August 28, 1998
9:30 AM
422 Snell Engineering Building

Abstract

A plethora of broadband and multimedia services are starting to become available to the residences. Formerly impeded by the limited bandwidth of the ubiquitous data modems, services like Video-on-Demand, tele-conferencing, remote gaming, fast information accessing, efficient voice telephony and many more are making their path to our homes, schools and businesses. An emerging shared medium network that is expected to provide this path to the residences, consists of a combination of optical fibers and coaxial cables referred to as Hybrid Fiber Coaxial or (HFC). The large bandwidth available in HFC networks leaves the cable industry with a major challenge as to how to best exploit this bandwidth for digital data communications. In a HFC network, a centralized controller controls the bandwidth allocation to a plurality of end users. A MAC protocol is required such that the end users must observe a common set of rules in order to arbitrate transmissions on the shared medium. One MAC protocol has been specified by a consortium of cable operators, known as MCNS (Multimedia Cable Network System). The MCNS specification does not describe how transmissions are to be scheduled and therefore, the design of such scheduling algorithms is needed. Since effective scheduling decisions require knowledge of the system offered load - not available a priori - it is important that mechanisms be developed which estimate the offered load based on observations. This is precisely the problem addressed in this thesis. In particular, an offered load estimator is developed based on channel observations over some time interval. The performance of this estimator is investigated and is improved by introducing modifications which provide for better reaction to offered load variations. In addition, a computationally simple mechanism is proposed for testing the reliability of the observations and rejecting the incredible ones, enhancing further the performance of the estimator.

Thesis Committee:
Prof. Ioannis Stavrakakis (advisor)
Dr. Whay Lee (Motorola I.N.G.)
Prof. Ibrahim Matta (CCS)