Teletraffic theory for the Internet
-----------------------------------
Electronic Engineering & School of Mathematical Sciences
Queen Mary & Westfield College, University of London
The two departments have joined to form a collaboration with BT laboratories
to investigate and develop the teletraffic theory needed to model Internet
behaviour. In particular, the project aims to address the challenge of long
range dependence and the implications of this behaviour for performance
engineering, dimensioning, and control policies.
A CASE award for a PhD in Statistics is available as part of this BT funded
project on teletraffic theory of up to three students. The statistics
project will involve the modelling of web cache behaviour. The other PhD
studentships are in non-linear dynamics and electronic engineering.
Applicants should have a good honours degree in either an appropriate
discipline. Please note that applicants must satisfy EPSRC eligibility
rules. The CASE award funded by BT Labs, augments the basic EPSRC
studentship rate to give a minimum grant of not less than £12500 p.a. over
three years. During the project each student will have the opportunity to
work at BT Labs, Martlesham for approximately three months in total.
Applications forms are available from the Registry* at QMW. Further
information is available from Dr Barbara Bogacka (0171-975-5497) or Prof.
David Arrowsmith (0171-975-5464).
*Higher Degrees Office (0171-975-3751), Queen Mary and Westfield College,
University of London, LONDON E1 4NS.
http://www.qmw.ac.uk/
"WWW Cache Statistical Modelling"
---------------------------------
The project involves non-linear modelling of a cache's performance, where
long range correlation of observations is present [1].
Several cache metrics have been proposed and investigated in the computer
science literature. The most popular are hit rate, popularity function and
Hurst parameter [2], [3]. The dependence of the metrics on the cache
infrastructure parameters has also been considered. Mathematical models
have also been proposed allowing prediction of cache population dynamics,
[4]. Furthermore, some simulation studies have been performed in order to
find out the main factors governing cache behaviour, [5].
The subject needs further development, particularly in investigation of the
impact of long-range dependency on the planning of network infrastructures.
A multi-level cache model with parental and sibling links involving user
communities with different file popularity curves needs to be built and
tested on several representative sets of data. An important element of the
statistical modelling will be a full investigation of the consequences of
differing assumptions about the stochastic elements of the model. The work
will need to cover the characterisation of variability and the estimation
and testing of those model parameters suitable for the diagnostics of cache
performance.
It may be necessary to introduce new metrics, which could be very useful to
the Internet researchers for cache planning. A detailed analysis of the
proposed metrics would have to be performed.
A successful PhD student should have very strong computational
(programming) skills as well as good statistical knowledge.
[1] Beran, J. (1992) "Statistical Methods for Data with Long-Range
Dependence", Statistical Science, vol. 7, No 4, 404-427.
[2] Roadknight, C., Marshall, I. And Vearer, D. (1999) "File Popularity
Characterisation", Technical Report (?).
[3] Marshall, I., Boissaux, M. and Roadknight, C. (1998). "Periodicity and
Self-similarity of Web-cache traces". USITS Symposium (?)
[4] Bilchev, G., Roadknight, C., Marshall, I. and Olafsson, S. (1999).
"WWW Cache Modelling Toolbox", The 4th International Web Caching Workshop,
San Diego, California, March 31-April 2, 1999.
[5] Abrams, M., Standridge, C.R., Abdulla, G., Williams, S. and Fox, E.A.
(1995). "Caching Proxies: Limitations and Potentials". Proc. 4th
International Word Wide Web Conference, Boston, MA, December 1995.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|
