Many data communications protocols allow variation of certain
properties, such as frame length, to adapt the performance
to a changing (non-stationary) communications environment.
However, this is often a one-time setup (stationary) and any
changes of the environment lead eventually to a degraded
performance, if not to a total communications failure. Some
research results are available for the case where the frame
length is adapted to the measured error rate to achieve a
rudimentary single-variate adaptation method for simple
stop-wait (idle RQ) protocols.
Our research focuses on general adaptation possibilities
of existing protocols to the communications environment
(what to adapt ?). Rather than simple stop-wait protocols,
more common sliding-windowing protocols are considerd which
allow for a multi-variate adaptation of several protocol
attributes, such as frame length, group size, window size
(also number of windows for multi-window type protocols),
and retransmission mechanism.
A strategy of adaptation is to be developed so that the
protocol varies its attributes to give the best performance
with changing environments subject to a stability criterion.
The conditions to use as adaptation criteria are to be
determined (when to adapt ?), and the best combination of
adaptable protocol attributes (what to adapt ?), and adaptation
algorithms (how to adapt ?) are to be found.
First, a simple analytical model is to be developed which
can beneficially be used to gain insight into the complexity
of multi-variate adapation, and hopefully leads to a conceptual
guideline for the best performance optimization approach.
Subsequently, the adaptive protocols are to be simulated
in a software testbed for validation purposes. This allows the
testing of the selected procedures by simulative, yet practical
experiments. The performance measures appropriate for the
adaptive process (e.g. throughput, channel utilization,
adaptation speed, stability, optimality) are to be measured.
Participating in the research are Dr. Bernd J. Kurz (professor,
CS at UNB), Dr. Larry Hughes (professor, St. Mary's Univ., Halifax)
with the assistance of MCS graduate student Manvinder Singh.
Project Status: Started: May 1997
Completed: May 1999
"Multi-Variate Adaptive Protocols for Data Communications by Manvinder Singh, MCS thesis, Faculty of Computer Science, University of New Brunswick, May 1999
Back to: Kurz's Research Page | Kurz's Home Page
Last revised: 11 March 2004 by Bernd Kurz bjkurz@unb.ca