Parallel Modelling, Simulation and Performance
Analysis of Discrete Event Systems
- The field of Discrete Event Systems (DES) has traditionally been
connected with sequential discrete simulation. In the last decade,
there has been increasing interest of system theorists for "discrete
event systems", whose evolution in time is marked by the
occurence of events, and of computer scientists for
nonprocedural, parallel and distributed approaches in simulation
processes. Tracking and linking of these directions may bring
benefits for both system theory and computer science.
- In my doctoral thesis I propose a systematic approach of Discrete
Event Systems field, trying to link the main results and to extend
some of them.
- The main topics I am concerned with in this stage of thesis
elaboration are:
- 1. The questions of representation and parametrisation
(modelling), together with the algorithms to go from one of these
models to another and the significance of various system
properties in each of the possible representations.
- 2. The suitability of these models for performance analysis and
optimization of DES.
- 3. Parallel and distributed simulation - as an adequate
performance evaluation tool. Traditionally, simulations of
complex discrete event systems have proved to be exceedingly
slow. Using multiple processors for these simulations appears to
be a promising approach to improving their speed. Since most
simulations are of systems consisting of many components
operating in parallel, it seems reasonable to suppose that the
inherent parallelism in the system can be well-exploited by the
simulation.
- 4. Synthesis algorithms and specific design principles of such
systems.
In this stage I am also concerned in designing software tools for
parallel discrete event simulation and finding new applications for
them. An experimental conservative parallel general simulator has
been designed and implemented in two versions within the
Department of Computers at the University of Craiova. The initial
assumptions on which the whole design and implementation of
this simulator were based included the fact that the typical user of
the simulation system may not be aware of parallel algorithms,
parallel decomposition of systems or the way of efficiently charge
of the available processor network. It actually runs on a
transputer network and a new version, running in a cluster of
workstations, will be soon available.
- Last updated 1 November 1996
Back to main page...