🚀 go-pugleaf

                    Parallel Simulation of ATM Networks
                              Fabian Gomes
                  TeleSim Project, University of Calgary

Abstract:

The principle objective of TeleSim is to produce network and traffic simulation
tools capable of evaluating performance during ATM network design, development,
planning and  operation.  To address the computational  demands of  ATM network
simulations,  the project has developed  both sequential  and parallel discrete
event simulation executives.  Two parallel executives  based on  optimistic and
conservative protocols are presented alongwith their performance.

Two key related problems - efficient and transparent state saving in optimistic
parallel  discrete  event simulation  are investigated.  Several variations  of
incremental strategies  and their optimizations  are investigated.  Two schemes
for automating  incremental  state saving  are presented.  The first,  based on
language extensions,  uses parameterized types,  operator overloading,  and new
`recover'  and  `nonrecover'  type-specifiers.  The second,  a  compiler  based
scheme,  relies on static  data flow analysis  of programs  for efficient state
saving.  A 65% reduction in  state saving overheads  was noted using the  later
scheme in a telecommunication simulation.


Biography:

  FABIAN GOMES is a post doctorate fellow in the Department of Computer Science
at the University of Calgary.  He received his Ph.D. degree in Computer Science
from the University of Calgary in 1996.  His research interests are in parallel
simulation,  modeling ATM broadband networks,  distributed systems and rollback
based computing.