[Date Prev][Date Next][Date Index]

a PhD seminar in Network Research Group Seminar (10.00-11.00 am, Friday, March 23, 2001, in room MSCS315)



During the next meeting of COSC/EEE Network Research Group,

at 10.00 am, on Friday, March 23, 2001, 
in room MSCS315,

Mr Sven Ostring (a PhD candidate from the Dept. of EEE 
at the University of Canterbury) will conclude his PhD study 
by a seminar talk on

Reactive Traffic Control Mechanisms for Communication Networks
with Self-Similar Bandwidth Demands
====================================================================
Abstract:  
In this talk, the research work reported in the recently 
submitted EEE Dept thesis with the same title will be presented. 
There is significant evidence that network traffic is self-similar 
by nature, and there have been a number of studies which have 
investigated the impact of this characteristic on network 
performance.  In contrast, we have been interested in harnessing 
unique properties of self-similar traffic, such as long-range 
dependence, so that high performance network management algorithms 
can be developed.  Long-range dependence implies that traffic has 
higher predictability when one uses algorithms based on traffic 
measurements.  We have developed a predictive rate control algorithm, 
labelled PERC, for the ABR service, which uses the stochastic
structure of network traffic to allocate available bandwidth.  
An enhanced version of PERC, referred to as PERC+, will also be 
presented. The performance of these algorithms have been analysed 
and compared with other rate control algorithms using simulations.  
Other aspects, including the sensitivity of the algorithms to 
traffic parameters and solutions for robust performance, are also 
considered.  Finally, we consider the influence of short-range 
correlations on traffic management, as compared with long-range 
dependence, and subsequently determine methods of simplifying 
the algorithms while maintaining high performance.
====================================================================


Everybody welcome.