[Beowulf] SGI to offer Windows on clusters

Joe Landman landman at scalableinformatics.com
Sun Apr 15 16:26:49 PDT 2007



Jeffrey B. Layton wrote:
> Robert G. Brown wrote:
>> On Sun, 15 Apr 2007, John Hearns wrote:
>>
>>> And re. the future version of Scientific Linux, there has been debate 
>>> on the list re. co-operating with CENTos and essentially using CENTos 
>>> as a base, and SL being an overlay of specific application and 
>>> library RPMs.
>>> Pros and cons either way there.
>>
>>
>> IMO, most cluster builders will find it more advantageous to track the
>> FC releases instead of using RHEL or Centos or things derived therefrom.

  Only if they are not building clusters for commercial customers, or 
customers with specific OS (distro) requirements.  FC simply will not 
fly in a shop that demands long term support.  We deal with lots of these.

   I know there are many users and supporters of FC here.  I have 
used/played with 1,3,4,5,6 so far.  I won't comment on quality.  I will 
comment on longevity and perception.  FC is not designed to be a long 
term supported distro.  It is, by design fast moving.  It has, as a 
result of its rapid evolution, more bleeding edge bits than regular 
longer haul distros.  On the perception side, pretty much everyone 
(RedHat included) with a few notable exceptions on this list, views FC 
as less a distribution unto itself and more as a testing ground for 
future RH technologies.  This doesn't mean people haven't been using it 
as a distribution unto itself.  If you don't need and don't care about 
long term support, don't care about occasional ... surprises ... in 
upgrading (FC4->FC6 on one of my machines caused it to kernel panic 
almost immediately, until I added some magic to /boot/grub/menu.lst to 
stop this), it could be a fine distro for some groups of users.

>> Hardware support is key, and Centos can get long in the tooth pretty
>> quickly in a cluster environment with any sort of annual turnover.
> 
> I think this is an interesting question. On the one hand, you're absolutely
> right in that faster moving distros like FC (or Ubuntu) allow newer 
> hardware
> to be used quickly and newer kernel developments to be used to advantage.
> 
> On the other hand, ISV's, who many cluster users depend upon (not so many
> on this list) are reluctant, to say the least, to switch to a faster 
> moving distro.

Heh...  Few if any ISVs have ditched their RH7.3 support to date (or 
SuSE 8).  Years after such distros ceased to be marketed/supported by 
the vendors.

This is one of the strongest arguments to certifying to an LSB and not a 
distro.  Unfortunately, some distro makers (cough cough) choose to 
largely ignore the LSB.  Which means ISVs certify to distros.  Which 
gives us the current situation.

> They are happy with something like RHEL or SLES. I know one ISV that
> only recently switched to RHEL 4 from RHEL 3! There are many more
> ISV stories to tell (I'm sure Greg could share some as well), but believe
> me, they don't want to change and many times, won't change.

New distros and new releases of distros cost them money to certify 
against.  Sure, if their software were open, the certification could be 
public.  I somehow don't expect these ISVs to do that.

> Of course, in defense of the the ISV's, some of their customers are 
> required
> to keep certain versions of their software on certain OS versions for a
> very long time. So the ISV's have gotten very conservative in their choice
> and support of distros.

I count largely 2 distros supported these days.  RH of some flavor, and 
SuSE of some flavor.  Most everything else is in the noise (though 
Greg's company and a few others do support more).

> So this leaves us, and many times cluster vendors and integrators, in a
> pickle. If you need ISV codes but also want new hardware, what do you
> do? I don't have a good answer. The only answer I can think of is to 
> continue
> to educate the ISV's and perhaps get them to support something a little
> more basic such as a kernel/glibc combination or at least a very basic
> (i.e. minimal) set of "parts" in a distro.

LSB.  It is sort of like an ABI for MPI.  It really needs to happen. 
LSB support is at this moment, more of a checkbox on government 
procurements, as compared to being something that ISVs strive for.



-- 
Joseph Landman, Ph.D
Founder and CEO
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://www.scalableinformatics.com
phone: +1 734 786 8423
fax  : +1 734 786 8452
cell : +1 734 612 4615



More information about the Beowulf mailing list