[comp.benchmarks] Archiving results as well as source code?

morgan@ms.uky.edu (Wes Morgan) (04/13/91)

I've been maintaining an archive of benchmark source code at
wuarchive.wustl.edu (128.252.135.4) for about 2-3 months now,
and it has been reasonably popular (according to wuarchive's
statistics).  My thanks to everyone who has supplied source
code; anyone with additions is welcome to send them to me.

When I originally opened the archive, I decided not to archive
results.  This decision was based on several points:

   -- There is no guarantee that all results submitted to me
      have been generated under the same circumstances, i.e.
      single-user mode, native I/O devices (as opposed to 
      NFS/RFS resources, which could skew I/O benchmarks),

   -- I received *widely* variant results for the same make/model
      of system.  This is probably a result of subtle differences
      such as those described in the comment above.

   -- Many systems are available in single- and multi-processor
      models.  For instance, the AT&T StarServer E may include
      anywhere from 2 to 8 80486 processors.   Interpreting results
      from these product lines will be confusing at best and mis-
      leading at worst.
     
   -- People were sending me results transcribed from trade rags,
      sales literature, et cetera.  I don't feel that these reflect
      end-user installations; therefore, these results could only
      be used as a "best case" metric for that type of system.

If the demand is there, I am willing to archive results.  I would
request that the report include installation/configuration-dependent
information such as that mentioned above.  I will probably archive
the results as digests of electronic mail.  Of course, the entire
library of results would be a 'grain of salt' collection at best.

Discussion/comments/suggestions?

Wes

ps> My thanks to those who have been so generous with their advice.
    This is my first attempt a maintaining an ftp archive.  I'm always
    open to suggestions and/or comments.

-- 
 morgan@ms.uky.edu    |Wes Morgan, not speaking for|     ....!ukma!ukecc!morgan
 morgan@engr.uky.edu  |the University of Kentucky's|   morgan%engr.uky.edu@UKCC
 morgan@ie.pa.uky.edu |Engineering Computing Center| morgan@wuarchive.wustl.edu
    Curator of the benchmark archives at wuarchive.wustl.edu <128.252.135.4>