long reply (was RE: LONG RANT [RE: RHEL Copyright Removal])
laurenceliew at yahoo.com.sg
Wed Nov 26 17:56:53 EST 2003
It is great that you have convinced your management to use Linux and
winning the game :-)
you should look at using RHEL WS for the compute nodes and RHEL ES or
AS for the frontend. It will lower your costs quite a bit.
for 1 frontend with AS (USD1499) and 42 compute (42 x USD179) =
USD$9017. (for 42 nodes, u probably can and should get discounts!)
Your cost of your hardware would probable amount to around USD100K. So
the OS costs comes up to about 10%... I believe 10% for OS for a cluster
is about right.
I understand how you feel about RHEL policies etc, and I am hopeful that
RH will have specific HPC pricing further down the road.
I would encourage you to speak to your RH rep nicely and explain what
you are doing and why you think you should get a "HPC" pricing. (ie.
more discounts of the compute nodes)
You will be surprised that not all in Red Hat appreciates HPC and what
we do, and why their model of pricing currently does not work for us.
As for alternative distro, you may wish to look at Novel/Suse Linux and
use it as a counter balance to RH.
Again, you will note that I can only encourage the use of these
"commercial" distros as it will probably be part of a ISVs supported
matrix. Most of my customers are sticky about such support and demands
that the OS used is a supported OS for their applications.
I look forward to the day Novell Linux offers a HPC pricing model. But
again for them to do so and provide the support and patch stream, there
will be costs and I hope it will be reasonable and which something the
community can accept.
On Thu, 2003-11-27 at 01:48, graham.mullier at syngenta.com wrote:
> Laurence Liew wrote:
> > The cluster community have done very well and today, large commercial
> > organisations are adopting linux clusters as one of the tools they use
> > to solve their complex problems.
> Yup, that's us - we are currently adopting a variety of open source tools,
> including Linux, to help tackle at least some of our HPC needs.
> > But I find this talk of "stripping" RHEL copyright to create
> > yet another
> > distro to be counter productive as linux beowulf clusters goes into
> > commercial mainstream computing.... where customers have specific
> > support demands. (And yes... commercial customers WILL PAY
> > the full list
> > price of RHEL to build a cluster).
> > Now... I believe the USD25 and USD50 are acceptable pricing for the
> > value that RHEL + RHN brings to the customer (academic). The
> > cost of the
> > OS is a small fraction of the total value of the cluster.
> > Most of our users want a stable and supported OS, but more
> > importantly,
> > most of them run a commercial software of one form or another... and
> > this means that these 3rd party ISV softwares are most likely to be
> > certified on RHEL.
> I think you are confusing things here (I know you are ranting but let's try
> to keep the arguments coherent, please! ;)
> I'm running a project within a commercial company, so academic rates are of
> no use to me. I am willing to pay for what I get, but I'm not willing to pay
> simply to give us a warm glow that we are "supported". If I get some value
> I'll pay. I don't think I get value if I'm expected to pay separately for
> each copy of RHEL-AS on each of 42 compute nodes, and the only price I'm
> offered is an extreme full list price. I would be willing to buy into a
> model where I'm paying for a clean, well-tested patch stream. But that model
> can not scale cost linearly with number of installed nodes - I'm not even
> convinced it can scale as the log of the number of nodes.
> > if the community continues to fork a project just becauses it charges
> > some $$$$, our progress would be very slow.... Redhat have listened to
> > the customer and partners and have created a academic pricing
> > model for
> > cluster builders... so we should accept that and move on.
> As I've said above, this is simply confused and does nothing for me or my
> project. The community depends on people contributing work - and in some
> cases those people contribute work in exchange for remuneration. But in
> other cases we as a community find ways of driving development forward
> through what amounts to barter - we all get value from the open source
> software, and we all contribute to it in some way.
> RH is (or at least appears to be) going down the restrictive licence,
> over-priced model pushed by MS. They've also learned the 'force frequent
> upgrades' trick. That leaves me uncomfortable about them as a vendor with
> whom I believe I'll have a good long-term relationship.
> But in the short-term software I use needs "RH 7.1", or "supported only on
> RH 7.3" or "RHEL-AS 2.1". Great. So I want ways of using RH that reduce my
> risks (what if RH stop making binaries available - can I still operate? If
> not, I want to be able to recompile from the source, and need to avoid
> copyright infringement problems).
> > disarray and you will see droves of commercial ISVs
> > abandoning linux and
> > moving back to UNIX and Windows....
> > where would that leave us? without commercial apps, linux would never
> > sustain and grow in the commercial arena.
> ah, well, now you've moved off into another universe. This isn't the one I'm
> in. Closed source is bad - it gets in my way, makes my life difficult, and
> increases my project's risks enormously.
> Why should I pay RH huge sums of money for Linux AND have to fight to get
> acceptance of Linux internally when I could take the "easy" option and just
> buy Windows? [by the way, I know why, and I'm fighting - and winning]
> Where I am now is a small part of the commercial arena, it uses commercial
> apps that run on Linux because we, customers, demand that they do. If RH
> make life difficult for us (awkward licence model and/or high price per
> node) we will start looking for ways around the problem, because it is
> worthwhile. Maybe we'll shift to another distro, maybe we'll take the time
> and sort out how to build it ourselves - and once we've done that, what use
> are RH to us? And if they are no use, will they get any money - no I don't
> think so.
> Open source is a whole new way of working - and the money has to come in a
> different way. If we're offered useful services that we can't or don't want
> to handle internally, we'll look at buying them. But if the price is too
> high we won't bother.
> (long term IRIX user, computational chemist, and now chemoinformatics
> specialist. I put up with Windows for office use but wouldn't want to rely
> on it for anything important...)
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
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