[rescue] intel vs. sun- for real

[Joshua D Boyd]

On Fri, Jul 19, 2002 at 09:35:06AM -0700, Koyote wrote:
> Unfortunately, due to the institution in question and the need for support contracts, I can't sell them a spare ultra1. So no, the debate is still about *why* Dell/Intel would be a poorer choice than Sun.

Well, I wasn't suggesting that you actually sell them a U1.  I was
suggesting that you prove it with a U1, then sell them the minimum
decent (ie, non U5 or U10) machine with a significant amount of
support.  But if you can't do that, then you can't do that.

> Unfortunately, what my girfriend's boss is looking for is some
> evidence that someone in the world has done some testing on
> various architectures beyond religion. They don't get to do
> loaners- and honestly, the amount of time it would take the
> university department in quesiton to set up, learn to do, do, and
> analyze tests precludes that option. 

Well, I'm sure that Sun can point you to some info.  I haven't not
seen any independent tests with Sun vs. other things that have really
meant anything to me though.  I've seen some interesting linux version
a versus linux version b tests, and some interesting linux versus some
bsd tests, but no good linux versus solaris tests.

> See, This is where the debate breaks down. I can go test how much
> packet I can push on a u2 with a qfe, yes. But stating that there is
> no valid data out in thw world at all is a bit ... well, it's her
> boss, not mine. She has to give the answer. 

I don't know what to say.  Maybe sun or others can help, but I just
haven't seen any good Solaris tests.  

People on the list have talked about starting a company.  Well here is
a company idea.  Do meaning full performance tests of numerous
platforms, and compare them in various ways.  In particular, all
different, web, ftp, samba, and NFS loads, high memory loads, huge
task count loads, etc.

But to get back to your problem, one meaning full thing you could
point out is that there are some tests showing that Solaris has a more
efficient TCP/IP stack (aren't they the ones with the zero-copy
stack?).  This really wouldn't demonstrate that Solaris is definately
better, but it is a step in the right direction.

-- 
Joshua D. Boyd