Mainframes vs clusters

[Mike Coleman]

Jonathan Hutchins <hutchins at opus1.com> writes:
> From: Mike Coleman [mailto:mkc at mathdogs.com]
> > I was quite shocked to discover
> > that the laptop was the faster of the two, by a significant 
> > margin (maybe 2x, I forget).
> 
> But were you really running an equivalent environment on each, ie plain
> single-session console with no servers or deamons in the background (or at
> least the same/equivalent ones)?  

Yes.  Well, since one machine was running Solaris and the other was running
one of Debian/Red Hat/Slackware Linux, the daemons wouldn't be exactly the
same.  I'm pretty sure both machines were otherwise unloaded, though.

One possible explanation is that early versions of Solaris 2 were possibly a
bit slow/inefficient.  My wild guess is that it was due to Linux's speedy
network code.

> I'd be willing to bet that you could do just about as well in an xterm
> window on the Sparc, while trying to load XWindows and KDE on the laptop
> would bring it to it's knees - but that could be the same kind of erroneous
> assumption:

The laptop only had 4 or 12 MB RAM, so running X on it definitely would bring
it to its knees.  The SPARC had a lot more memory (32 or 64 MB?).

I'm not saying that that laptop would beat out the SPARC 2 in *all* tasks.
Probably it wouldn't.  But my assumption that it wouldn't beat out the SPARC 2
for this particular (low memory, high network) task was wrong.

I've been similarly surprised with other obvious assumptions; e.g., SCSI is
always faster and more reliable than IDE, Linux will always beat the crap out
of Windows for webserving, etc.

-- 
Mike Coleman, mkc at mathdogs.com                                      Windows XP
http://www.mathdogs.com                                               Linux :)
problem solving, expert software development