Soft Protocols research in the FAN Club
Overview
The research is founded on a component-based view of computer
communication protocols, in which we have identified a dozen
largely-orthogonal underpinning principles that lead in turn
to corresponding protocol techniques. This gives a horizontal
perspective on communication systems, in contrast to the
vertical perspective encouraged by the OSI reference model of ISO.
Systems are still implemented in a layered manner, but there
are no preconceptions about which functions are handled in
which layers.
The work blends ideas from human communication, from formal
methods and from novel system architectures. The new perspective
is explained and justified in extended form (470 pages) in the book
"Computers in Communication"
which was published by
McGraw-Hill in 1997.
This research has application in our
flexible architecture for
networking applications research.
Areas of investigation
Active protocols
Active protocols are soft protocols used in an active network
setting. Active protocols are defined, constructed and used
dynamically, perhaps very dynamically on a per-connection basis.
In general, active networks allow arbitrary (safe) computation
at network nodes. Active protocols make use only of domain-specific
computation - the domain being communication protocols - and so
enhance reliability and safety. A prototype active protocol
system is being constructed, using implementation over an
"ABone" - an active protocol backbone.
Theory of soft protocols
This involves the specification and verification of
soft protocol behaviour, using the Calculus of
Communicating Systems (CCS), augmented by a new addition
operator over actions. This allows soft protocol components
to be specified, and their behaviour analysed. The longer-term
aim is to mechanise the analysis in a "protocol
workbench" - a tool for analysing communication protocol behaviour.
Modelling and simulation of soft protocols
This involves
hierarchical methods that model system components at the coarsest
level of detail adequate to obtain valid results, together with improved
simulation algorithms and the use of supercomputers to assist
with brute-force simulation. Validation is carried out as an
interative process, to find the most efficient hierarchical
model structure for production-run simulations.
Back to FAN Club page
Gordon Brebner, questions to <gordon@dcs.ed.ac.uk>,
26 March 1999