Cluster Poll Results (tangent into OS choices)

Joe Landman landman at scalableinformatics.com
Tue Nov 4 07:38:30 EST 2003



Mike Snitzer wrote:

>On Mon, Nov 03 2003 at 21:58,
>Joe Landman <landman at scalableinformatics.com> wrote:
>
>  
>
>>The "smile" test?  I thought it was all about risks, support, etc.  
>>ROCKS appears to be in significant use as indicated by the ROCKS counter 
>>page.  
>>    
>>
>
>In my _personal_ utopia of the industry smile-tests are worthy; I do
>  
>

Ahh...  so thats what I need to win more business.  The smile test...

(...)

>>Remember that RedHat's added value is in packaging, bug fixes, etc.  
>>    
>>
>
>Not to mention numerous contributions to the Linux kernel, low-level
>libraries (nptl), compilers and much more.
>  
>

This was specifically implied/mentioned in my post.  "They bundle many 
peoples' code (Don's and probably a number of others here).  They have 
added value back to the community as a whole."

>  
>
>>I would argue that Fedora is more like a permanent beta.  It doesn't 
>>look like we will get good things into Fedora anytime soon (x86_64, XFS 
>>et al), and the release/support cycle is too short to be useful for long 
>>term customer support.   The risks of that platform would be somewhat 
>>high for a commercial deployment, and I would find it hard to justify 
>>installing this for a customer knowing full well that next year, they 
>>are support free.
>>    
>>
>
>It all comes down to opportunity cost; time spent working with the Fedora
>project (and its evolving policies) to add required features is 
>time consuming and takes away from _real_ HPC innovation. 
>

Yes.  That is the point after all, that if you spend all your time 
discussing whether or not your masters^H^H^H^H^H^H^H benefactors will 
accept XFS or other HPC relevant features in the kernel, you do not then 
have time to put them in.   Part of it is opportunity cost, the other 
part is a zero sum game of time.


> BUT, if the
>entire HPC community actually worked together to bring about that change
>it wouldn't be that hard.  Too idealistic?  
>

I believe it might be too idealistic.  This crowd, if you read this 
forum and some of the others, likes to innovate and create its own value 
atop some sort of standard offering.  If I am reading you correctly, you 
are advising focusing on making on particular platform that you 
personally (to separate you from your employer here) like, as the 
standard, and stop all the bickering about doing another direction (that 
you personally do not like).  Is this a  fair read?

>It would appear so based on
>the resounding cry for rebuilt RHEL solutions.  Keep in mind that
>customers want "the real thing".
>  
>

Hmmm.   The one thing that customers repeatedly tell me is that they 
want their solutions to be supportable.  They don't want one-offs, or 
other things that significantly increase their risks.  If the "real 
thing" represents a risk, they will not go for it.  If the 
"non-real-thing" ala ROCKS, Warewulf, CLIC, OSCAR, et al. are well 
supported, and slowly varying enough, they are content to live with a 
few warts. 

More important than this is a specific understanding on the part of the 
customer that RedHat is focused upon a different market.  As I noted in 
the post to which you replied "That said, they are not terribly 
interested in HPC from what I can see.  Might be due to the size of  
this market compared to their total addressable market. "  They, in this 
case, are RedHat.  The "Real-Thing"(TM) doesn't matter in this case, if 
it is missing key functionality/features et al. 

Moreover, Fedora will be no-more the real thing than the RHEL based 
versions.  Though, the RHEL will vary more slowly over time than Fedora, 
which is a very good thing for a stable commercial/academic cycle shop.



>  
>
>>Ripping of RedHat?  I thought they were packaging GPL and similar 
>>software... how is taking GPL software which is Libre' and 
>>redistributing recompiled versions of it (allowable under the license)  
>>ripping off the folks who have a their own packaging of it?
>>    
>>
>
>It comes down to the unfortunate reality that many in the HPC community
>would rather continuously fork/reinvent RHEL than work with Red Hat to
>arrive at a mutually beneficial arrangement.
>  
>

As noted previously "That said, they are not terribly interested in HPC 
from what I can see.  Might be due to the size of  this market compared 
to their total addressable market. "

This market is not significant to them.  It will not drive hundreds of 
thousands of additional unit sales.  It requires levels of support that 
they may be unwilling to supply due to the individuality of the products 
offered.  Aside from that, finding and paying for real HPC people (e.g. 
more than a few years experience) is not cheap/easy.  This increases 
their marginal costs without really increasing their marginal utility.

I don't blame them for this.  It is simple economics.  It leaves a 
market hole that (as Glen pointed out with his company), people are 
willing to step up to fill.


>  
>
>>RedHat is focused upon its primary market, which appears to be 
>>Unix/Windows server displacement.  Mike's employer is focused upon 
>>selling hardware.  Glen's company is focused upon good quality cluster 
>>software.
>>    
>>
>
>While I appreciate you associating myself and my views with my employeer I
>have expressed my _personal_ views.  However, your assessment of my
>employeer's focus is not accurate; but I'm not going to get into that
>discussion.
>
>  
>

Highly oversimplistic assessment on my part.  I assumed that someone 
writing from an lnxi.com email address would be expressing corporate 
philosophy.  This is your own _personal_ opinion then?

>>For companies like mine, the issue is a stable reliable platform to 
>>build our product offerings.  The problem with things like the permanent 
>>beta cycles of Fedora is that we will have to focus more upon the 
>>underlying issues of the platform changes (which will not be focused 
>>upon HPC needs) than on our own development.  This is a moving target.  
>>This is "Not A Good Thing(TM)".
>>
>>A whole bunch of commercial software vendors have "old" and  "outdated" 
>>OS support for their wares.  I have to carefully check the software OS 
>>support matrix when building engineering or bioclusters.  RedHat 7.3 is 
>>long in the tooth, and it happens to be a very good cluster 
>>distribution, in large part because so many commercial codes have been 
>>ported in the RH7.x time frame.  
>>    
>>
>
>Make no mistake about it, its not good for any commercial company that
>historically relied upon Red Hat Linux; hence the extensive attention this
>debate has recieved all over the Internet.  You have blantantly attempted
>to spin this thread in a self-serving/tangential direction of company<foo>
>vs company<bar>; and it wasn't about that.  Now I know why this list is
>  
>

Uh... no.  Not even close, and mind the defensive bit, no one was 
attacking you.  There is no "spin" here.  Commercial software developers 
need stable bases for their products, pure and simple.  Nothing 
self-serving about that.

I didn't state whether or not I worked for a clustering company.  Though 
you asked us above to consider your opinion to be a personal one, this 
statement ties your posts to your company (IMO), and using your words, " 
attempted to spin this thread in a self-serving ..."  Glen gave his 
point of view, which I am guessing reflects his corporate positions.  If 
you are saying that Glen's statement is self serving, I guess I am at a 
loss to understand how yours is not.  I gave my point of view, and yes, 
it does reflect my company's philosophy.  I do not believe it is "self 
serving" to state what we perceive as product developers. 

>perdominantly technical and _tries_ to stay away from the commercial
>interests of any one vendor.
>  
>

I may be (slightly but purposely) dense at the moment, but I would 
suggest that you might wish to place a disclaimer in your posts as to 
when you post here, that these are your views and not those of the good 
folks at Linux Networx.   Otherwise, folks like me, Glen, and a number 
of others might make a mistake and assume you are posting their views.

>Mike
>
>_______________________________________________
>Beowulf mailing list, Beowulf at beowulf.org
>To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
>  
>

-- 
Joseph Landman, Ph.D
Scalable Informatics LLC,
email: landman at scalableinformatics.com
web  : http://scalableinformatics.com
phone: +1 734 612 4615


_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list