TI Monitoring Tools for Parallel Systems 

AU Heidi W.H. Tang 

RT MSc thesis 

LT UMCS-87-12-3 

OR UMCS 

AV http
http:www.cs.man.ac.ukcsonlycstechrepAbstractsUMCS-87-12-3.html 

AV email techreports@cs.man.ac.uk 

MN December 

YR 1987 

AB Monitoring is essential to assess the performance of a
computer system.   Parallel machines are fundamentally different
from conventional machines,  and may contain a large number of
concurrent activities.  Due to their  inherently complex nature,
a significant amount of experimentation is  necessary before they
can be built in a cost-effective and viable manner.   Therefore
new tools and techniques are needed to evaluate their
performance,  and to aid in their development and analysis. _
    This thesis describes monitoring computer systems in general,
and presents  three experimental parallel machine projects
currently under research in  the University of Manchester: the
ParSiFal, Flagship and Dataflow projects. _
    Two monitoring tools have been designed and implemented in
the context of  the Manchester Dataflow system, both tools
present monitoring data as a  graphical, animated display.  A
program level tool displays a dataflow  program executing on a
parallel machine.  Although the tools have only been  used on the
Manchester Dataflow system, they have been designed to be as 
source-machine independent as possible, and they should have
wider  applicability.