[Gllug] Recommendations for servers

j.roberts j.roberts at stabilys.com
Wed Nov 24 22:23:15 UTC 2010


On 24/11/10 19:24, t.clarke wrote:
> Interesting comment re ASUS m/boards and at odds with my experience.
> I have a server running for 10 years non-stop on an Asus m/b and another
> similarly for 5 years.   Maybe I am just lucky ?
>
> (the boards in question are server-grade motherboards, not the cheaper ones)

With the greatest respect (really :-)

- that's the problem with case studies.

Most ASUS server boards that we bought (hundreds, but less than 
thousands) did not fail.

However the failure *rate* was twice that of Gigabyte boards and five 
times that of Supermicro. We stopped using ASUS boards in 2005.

The lowest rate of all has been HP servers - so far 0% failure in 
service (in hundreds over many years: I'm not counting HDD failures, 
many of those!).

Umm - that's why we supply them to our clients - and swap them out at 5 
years or so. Of course that 0% in service fail rate could change tomorrow.

'HP business desktops' (and their dreadful consumer offerings) is 
something else entirely - we have had a whole batch of two year old 
machines start to fail.

I have myself two old Pentium 2 @400/550 machines with Intel 440BX 
chipsets built in 1998 or so, one with a Gigabyte and one with an Asus 
server board, in use in Spain since shortly after construction and used 
for years now as router/firewalls (running ancient Red Hat and Freesco). 
They've never failed, I'm wondering if they ever will! But that's a 
*case study*, and not a *rate*.

The assertion was that there was something special about a DNUK server 
hardware that justifies the price, (as against the Linux-friendly 
support package) and I am not entirely convinced. I'm not in any way 
saying that they are bad; and no doubt they maintain stats on their 
boards. But this is necessarily a retrospective view.

Concerning servers, the build quality of HP is superb; the build quality 
of Dell very good (usually); Supermicro components are excellent but 
system reliability will depend on the build quality and associated 
assembly line QA. I'd trust a Supermicro server more than a Supermicro 
serve board in someone else's box, say. I'd *nowadays* only build a 
server for myself, as a last resort - the component cost of a self 
assembled server usually exceeds the cost of buying the assembled item!

HP and Dell servers are built and tested in tens of thousands so any 
problems show up pretty quickly. HP supports Red Hat across its range. 
Dell is more problematic - but sometimes much cheaper.

But - I'm not making a recommendation, except for consideration.

MeJ
-- 
Stabilys Ltd		www.stabilys.com
244 Kilburn Lane
LONDON
W10 4BA

0845 838 5370
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list