🚀 go-pugleaf

In <4kiijh$9a2@rs18.hrz.th-darmstadt.de> kraemer@rzri6f.gsi.de (Michael Kraemer) writes:

>In article <danpop.829148336@rscernix>, danpop@mail.cern.ch (Dan Pop) writes:
>> If you're willing to pay the same amount of $$$ you're actually paying to
>> IBM for this kind of hardware support, I'm willing to bet you can find
>> a shop offering similar support for PC hardware.  The point is that you
>> can actually save a lot of $$$ taking a different approach with the PC:
>> with minimal technical skills, get a full set of spare components and
>> fix it yourself, way before you will even get a call back from the
>> service shop.  Without minimal technical skills, buy a spare machine,
>> similarly configured to the production one.  In case of problems, swap
>> the machines and send the broken one to be fixed.  This is still
>> considerably cheaper (for PC hardware) than a hardware maintenance
>> contract (for an RS/6000).
>> 
>
>If this would be true then why does your organization (CERN) run a 64(or so)-node
>SP2 instead of a pile of Linux boxes ? Or is it your boss who doesn't listen
>to your smart analysis ? 

For many good reasons.  One of them being that it would be hard to replace 
a parallel system by a bunch of Linux boxes.  However, PC's are being given
more and more attention here, as potential batch servers.  Their 
price/performance ratios as well as their absolute performance on CERN
benchmarks make them an interesting and cost effective solution.

>Moreover, from my experience it is much harder to get PC stuff fixed than AIX
>boxes. With an IBM contract it's "Dial IBM" and a hardware problem is fixed
>within 1 day. 

With a PC and a qualified technician is a 1 hour job.

>Of course software is much harder to deal with, as we all know.
>But who do you call for Linux support ? 

Any company offering Linux support contracts, e.g. Cygnus Support.
In most cases, the available documentation and the source code are enough
for a qualified person to fix the problem without external support.

>Commercial support might be more expensive, but on the other hand it saves me
>a lot of time I can spent on my main job, 

Lucky you.  When I had a serious problem with one of the components of the
batch system on the SP/2 machine you mentioned, I had to rewrite it from
scratch.  Later, IBM sent me the sources of their program and asked me to
fix their code, as well (in the frame of a joint project between CERN and
IBM).  Isn't software support great?  :-(

Dan
--
Dan Pop
CERN, CN Division
Email: danpop@mail.cern.ch 
Mail:  CERN - PPE, Bat. 31 R-004, CH-1211 Geneve 23, Switzerland