Re: Power supply efficiency on older model Sun equipment

On Mon, 1 Sep 2008, D G Teed wrote:

> The Netra T1 suggestion is a good one.  I've never
> worked with that size of SCSI drive before and had assumed
> they are very expensive.

One issue with the single-unit rack-mountable systems (and I see this all 
the time with my Cobalt's) is the fans. The fans are very small, so they 
have to work a lot harder to do their job, they have small bearings, are 
engineered well but just can't really tolerate long-term extended 
operation as the bearings fail a lot sooner that bigger fans with larger 
ball-bearing races, and the fans make more noise per unit volume of air 
movement than larger, generally slower, fans would.

That aside, for a while I was running the system which is currently an 
E450 as a Netra t105 system when it didn't need more than two disks. 
Worked really well, and in those the fans are bit better since they're not 
the ultra-thin type used in the Cobalt and similar 1-unit systems.

I think in terms of energy density, the 1-unit servers are really good 
from a scalability perspective as they're simple to set up, and are 
basically an Ultra 10 type system turned into a very well-made platform.

> Many of Robert's points are valid for the workplace,
> but not my situation at home.  It is always OK if
> I take perhaps 4-5 days to get my home server back up
> following a hardware loss.  But I try to reduce the
> chances of this by buying quality brand components,
> just as you would with the Sun.  Surely there are brands
> other than Sun which make good quality computing
> equipment.  Painting all x86 land with one wide brush
> isn't appropriate in my view.
>
> I believe the Intel and AMD (not yet retail) ultra low power
> CPUs (AMD also has a low power chipset) are the start
> of something that will hit the server room.  Sun will
> surely undertake the same initiative.  They may develop
> blade systems based on these procs, or go multi core, or
> somehow further improve on the original design.

I suspect the Opteron systems are intended to test out the methodologies 
of this approach.

> The benefit is massive.  Power rates always go up, so the
> savings just increase.  Any reduction in power used in
> the server room results in heat reduction and
> a double gain via A/C savings.

Absolutely! I have no A/C in my house and in older days when I was running 
stuff like Ultra 60's, and really early on, bit rack-mounted Sun-3/3x 
systems, they pumped out heat like crazy and on hot summer days the room 
was really hot. Before I bought this house and had all the older gear 
running in my mum's house (also had no A/C), it was worse since the 
ceilings were lower and rooms had much less air volume.

> It is partly an interest in low power computing which has
> helped to fuel my interest in the Intel solution.  The AMD
> chipset+CPU is 8 watts leaner than Atom, but isn't available yet or
> I would have gone that way.  Perhaps Sun will offer an AMD
> based solution like this?  But I've never regarded
> x86 Suns as true Suns.

They're not - for a long time Sun treated x86 as an after-thought and it 
wasn't until just recently in the last few years that Solaris for x86 was
're-invented' and given some significant development resources. I think 
the inroads Wintendo systems were making into the server market prompted 
much of that.

> I do like Sun servers, but brand loyality isn't as interesting
> to me as other things, like exploring the future of
> computing and saving my family some of the income for
> other things in life.

Indeed. Exactly what I'm trying to sort out at present while I've not yet 
landed a firm offer for a new full-time job.

Craig.

-- 
Post by Craig Dewick (tm). Web @ "http://lios.apana.org.au/~cdewick".
Email 2 "[email protected]". SunShack @ "http://www.sunshack.org"
Galleries @ "http://www.sunshack.org/gallery2". Also lots of tech data, etc.
Sun Microsystems webring at "http://n.webring.com/hub?ring=sunmicrosystemsu".