UNIX Performance Benchmarking

FRANEY jjf at hjuxa.UUCP
Tue Feb 23 03:12:43 AEST 1988


Does anyone have any leads or information
regarding UNIX performance measurement systems?

For purposes of market leverage, many computer companies
compare the performance of their processors to
the performace of their competitor's processors.
These comparisons have been sometime disclaimed as being misleading
due to hardware archetecture differences among computer manufacturers.

It seems that the industry may benefit from a common benchmark
system.  This system would be used as a common reference point
for performance measurement of the through-put of a UNIX system.
Marketing personnel would be able to say:
"Our UNIX complies with System V Verification Suite,
AND runs the System V Performance Benchmark in X amount of time."

Another purpose of a benchmark system would be for UNIX 
developers to compare the UNIX throughput on newly developed
machines.  These comparisons will satisfy hardware and software
developers that the new system is indeed X times faster than the old.

Another purpose falls into the realm of UNIX Multiprocessing.
It would be nice to know exactly how much faster
is a machine in multiprocessing mode (multiple processors enabled)
than the machine in single processing mode.

If this system doesn't already exist, it ought to.
If this system (or anything like it) does exist,
I request information on it.

If anyone knows of any systems which measure UNIX System V
performance, please email any information or contacts you care to.
I am also interested in hearing about any work being done in this area.
(I know that there is a POSIX working group on performance testing
but its status is unknown to me.)

Thanks
John Franey

(The ideas presented within this article may or may not be shared
by Digital Equipment Corporation.)



More information about the Comp.unix mailing list