Re: Windows v Unix Performance on Oracle 9i

From: Joel Garry <joel-garry_at_home.com>
Date: 6 May 2004 15:29:12 -0700
Message-ID: <91884734.0405061429.59a8290e_at_posting.google.com>


"Jim Kennedy" <kennedy-downwithspammersfamily_at_attbi.net> wrote in message news:<Cqhmc.37933$Ik.2505190_at_attbi_s53>...
> "Joel Garry" <joel-garry_at_home.com> wrote in message
> news:91884734.0405051441.6993253_at_posting.google.com...
> > subs_at_robtudor.co.uk (Rob) wrote in message
> news:<cd6e7d58.0405050615.5947cd17_at_posting.google.com>...
> > > Does anyone know if there are any good comparisons between Windows and
> > > Unix performance? Or does anyone have examples of Oracle databases
> > > running large numbers (200+) of concurrent users. I am trying to
> > > investigate what size implementations can be placed onto Windows 2000
> > > / Windows 2003 as opposed to UNIX but cannot find much data.
> > >
> > > Any information would be much appreciated.
> > >
> > > Cheers.
> >
> > As a matter of fact, yes:
> >
> http://groups.google.com/groups?hl=en&lr=&ie=UTF-8&oe=UTF-8&th=511ddc260aead7b6&rnum=2
> >
> > I also would be interested if anyone is running 200+ actual concurrent
> > users (not just connections) on Windows. I would find it hard to
> > believe.
> >
> > jg
> > --
> > _at_home.com is bogus.
> > Like using
> http://www.autoweek.com/specials/galleries/scaglietti/pages/612_Scaglietti_04.htm
> > as a school bus.
> I think you are going to get better scalability on Unix over Windows. I did
> work for a company and we had a very efficient application (bind variables,
> cursor reuse - parse once execute many) and were able to get many more than
> 200+ concurrent users. There are a ton of variables.
>
> How efficient is your application? (scalable)
> Is it more OLTP or OLAP? (ours was OLTP by far)
> What is the application doing in some sort of work flow for the application?
> How many disks?
> How much RAM?
> ...
> We were able to load test up to about 1,500 concurrent users on Oracle 8.0
> with our application on a very beefy Windows NT box. On Unix we were able
> to get a lot more users; our biggest problem was getting enough client
> machines so we could test UNIX!
> Jim

Thanks Jim, I now believe it. In that time frame I wouldn't have even tried it! I'm still not seeing the performance I would expect given the hardware, Dual Xeon gigawhateverhertz versus slower multiprocessor hp's, but the price differential... and what I don't know about Windows would fill many books... and grid is pushing linux... and I just haven't personally been able to make any really valid comparison tests...

The real thing I am troubled by is the idea that commodity hardware with no good diagnostics and uncertain quality control will affect databases in unpredictable ways. For example, I have a barely used W2K OAS10g mid-tier machine that locks up often in the middle of the night, for no apparent reason. Is there a hardware problem? A Windows configuration problem (those gazillion updates, what do they really do)? Does apache make too many files for its poor little file system? Is there something in Welchia or Sasser that Symantec hasn't figured out yet? How can I know?

jg

--
_at_home.com is bogus.
William Herschel, who discovered Uranus, wanted to name it "Georgium
Sidus" after the British King George III.
Received on Fri May 07 2004 - 00:29:12 CEST

Original text of this message