Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!uunet!zephyr.ens.tek.com!tekgen!tekcae!moiram
From: moiram@tekcae.CAX.TEK.COM (Moira Mallison)
Newsgroups: comp.databases
Subject: Re: Benchmark between dBASE & Oracle!
Message-ID: <5807@tekgen.BV.TEK.COM>
Date: 12 Apr 90 18:51:32 GMT
References: <1990Apr6.194145.2101@pds3> <509@dbase.A-T.COM> <1990Apr10.143336.2803@oracle.com>
Sender: news@tekgen.BV.TEK.COM
Reply-To: moiram@tekcae.CAX.TEK.COM (Moira Mallison)
Distribution: usa
Organization: Tektronix, Inc., Beaverton,  OR.
Lines: 37

In article <509@dbase.A-T.COM> dveditz@dbase.A-T.com (Dan Veditz) writes:
>>The Oracle license agreement does not allow buyers of the product to 
>>publish benchmark figures.  It ticks-off *our* marketing department!

In article <1990Apr10.143336.2803@oracle.com> tgreenla@oracle.UUCP (Terry Greenlaw) writes:
>Butt protecting personal disclaimers aside, of course Oracle wouldn't want
>competitors publishing benchmark results. The benchmark results we publish
>are all validated by independant firms that are a lot more likely to give
>honest answers than our competetors. 
>
>I'm sure Oracle marketing will be glad to provide you with all the 
>benchmarking numbers you would ever want....

I would trust numbers that come from a vendor's marketing department
about as much as you trust numbers coming from a competitor.  

As I understand the restriction regarding publishing of benchmark
results, it is not limited to competitive vendors.  It also extends
to independant companies that may want to implement benchmarks
in more than one DBMS, and report the results.   

I've done some benchmarking research work, and the number of variables
to control in order to get any meaningful results are greater than 
one might imagine.  Working in a tightly knit team with frequent 
communication about the implementations, it still took multiple 
walk-throughs of the code to insure that the implementations were
identical (ie that the benchmark operations were indeed measuring
the same database actions).  You just aren't going to get that
kind of consistency measuring Ashton-Tate's marketing numbers
against Oracle's marketing numbers.  

Certainly your application is the best measure there is of the
performance you can expect, but who has time to implement an
application multiple times?

Moira Mallison
Tektronix, Inc.