Re: Performance Measurements

Mike Curtin (mtc@research.megasoft.com)
Wed, 14 May 1997 11:41:28 -0400


Ethan M. O'Connor wrote:
>
> At 10:50 AM 5/14/97 -0400, Eric Z. Ayers wrote:
> >
> >2^26 complementary pairs of keys tested: 35 seconds (approx. 3800k keys/sec)
> >
> >
> >I wish I could use this cluster for deschall, but it doesn't belong to
> >me. There are 24 of these cpus sitting around (mostly) idle on an
> >internal lan somewhere in corporate america...
> >
> >-Eric.
> >
>
> I'll just note here that a single one of these Alphas would do
> 315997*2^20 keys/day; the 24 together would do 7583451*2^20,
> which would put them at sixth in the rankings. Yikes :)
>

Quick note: An SGI Power Challenge Array can scale up to 288 R10000
processers. Let's see:
(194 MHz R10000) 1.943M keys per second * 3600 * 24 * 288 =
48,348,057.6 million keys a day or 48.3 trillion keys a day
That is psu.edu and uiuc.edu combined! Does anyone have any connections
at SGI? :)

-- 
Mike