[rescue] A Question about Snap Servers

[Jonathan C. Patschke]

On Sat, 31 Jul 2004, Bill Bradford wrote:

> I'm all up on the "horsepower versus power consumed" thing lately for some
> reason...

There is that.

> I mean, I like the lunchbox sparcs, they're cute, but for the same or less
> power and in the same or less amount of space, I'm building a 1Ghz Linux
> box (or maybe Solaris x86, if it will work with the onboard ethernet).

And it still won't route packets any faster.  It'll only serve up
nameserver requests marginally faster.  Also, it'll require me to fetch
a keyboard and display should it ever decide not to boot, whereas the
SPARC will happily answer to the console server.

Now, if I wanted to use the box for any sort of computation[0], yeah,
more processor umph makes sense.  But, look at this:

[jp at sentinel:/home/jp]$ top
load averages:  0.23,  0.12,  0.09                                   23:41:43
33 processes:  1 running, 32 idle
CPU states:  1.4% user,  0.0% nice,  1.6% system,  0.8% interrupt, 96.3% idle
Memory: Real: 13M/28M act/tot  Free: 1028K  Swap: 61M/257M used/tot

Yeah, it needs more memory. :)

> I dont think i'll ever replace ohno.mrbill.net with a Linux machine,
> however. :)  Having 4M L2 cache on each of four CPUs is just nice.

Yummy.  Xeons?


[0] And, I will, soon, which is why it's about to be replaced by an
     Ultra 5.  The 'classic will probably become a fulltime nameserver as
     a result.
-- 
Jonathan Patschke )"We're Germans and we use Unix.  That's a combination
Elgin, TX        (  of two demographic groups known to have no sense of
USA               ) humour whatsoever." - Hanno Mueller in de.c.o.u.p