Unpaid deployments

Posted Nov 13, 2009 18:13 UTC (Fri) by jspaleta (subscriber, #50639)
In reply to: Unpaid deployments by corbet
Parent article: Notes from the LF End User Summit

Growing at similar rates... doesn't actually say anything about the nature of relationship between the two. Are the paid deployments driving unpaid or is it vice-versa.

Is the unpaid to paid deployment for linux distributions a close parallel to parasite and host relationships in biological ecosystems? And if so which is the host and which is the parasite? And what is the nature of the relationship? Is it primarily symbiotic where both groups benefit? Is it parasitic where one group is harmed at the benefit of the other? Or is it mutual convenience where for the most part neither is harmed or helped by the other (until an environmental stressor puts one community under pressure and throws off the dynamic balance)

I think we all want to believe the nature of the relationship its symbiotic. But I'm not sure we have metrics which point out that it is.
I certainly want to believe that non-paid deployments are beneficial long term. But then again, I'm probably part of the parasitic community and its difficult to convincing tell myself I'm causing harm for not paying for my deployments. As a species..we seem to have an inexhaustible capacity at denying our own accountability for harm if given half a chance to rationalize it away.

-jef

Unpaid deployments

Posted Nov 13, 2009 19:51 UTC (Fri) by drag (guest, #31333) [Link]

I would think that most places that do paid deployments also have unpaid
deployments for mitigating costs.

Many businesses require paid support for contractual reasons, or as part of
support package for applications, or for regulatory reasons. I think that
some places have requirements that even if they are capable of supporting
their own servers, they always hire out for support; there own admins are
not allowed to touch the hardware.

So there are all sorts of reasons why people do support, but they paid the
premium support costs for every little thing they want to use Linux on then
that would be massively more expensive then just sticking with Unix or
going with Windows.

As far as development goes.. for busy projects you really do need full time
people to at least do coordinating and documentation, even if most of the
development happens from third parties working part time.

So I expect it's heavily symbiotic.

Maybe when the market for Linux stagnates you might see some adversarial
stuff going on, but as long as it all keeps growing then everybody should
be happy.

Unpaid deployments

Posted Nov 13, 2009 20:56 UTC (Fri) by Cato (guest, #7643) [Link] (4 responses)

Even unpaid deployments create an ecosystem of skills, courses, books,
forums, blogs, etc - this is valuable to the paid-for product as long as
there's still a 'conversion rate' from free to paid-for that is high enough
(in absolute numbers) for the paid-for product provider to make a profit.

In some ways CentOS is a perpetual 'free trial' for people who are more cost
sensitive or want to deploy a new system very quickly, vs. people who need
commercial support - when some CentOS users get more budget, maybe in a
different company or somewhere that rapid support is worth paying for,
they'll go RHEL. If CentOS was very different to RHEL this switch would be
harder, and ultimately CentOS would be less valuable to Red Hat.

Unpaid deployments

Posted Nov 13, 2009 21:13 UTC (Fri) by jspaleta (subscriber, #50639) [Link] (3 responses)

Yes... very plausible..rationale... arguments about how the market value could be increased by unpaid deployments. All of which is completely plausible... but no metrics which quantify or support the theory. Just because its plausible and rationale doesn't mean we should assume it to be true.. human behavior isn't particularly rationale. I'd like to see some trendable metrics which attempt to show that the market value of the ecosystem is being driven by unpaid deployments in the way you theorize. Metrics...hard numbers...with error bars...can't forget the error bars.

-jef

Unpaid deployments

Posted Nov 14, 2009 0:27 UTC (Sat) by drag (guest, #31333) [Link] (2 responses)

Nobody keeps track of numbers like that.

It's nearly impossible to tell how many Linux vs Windows vs Unix servers are sold or being used.

Here is the problem:
Businesses don't advertise their IT infrastructure. They keep it secret because, frankly, it's nobody else's business. So the only way you can get a feel of a market wide is through server sales revenue.

Why? (you may ask)

Because most OEMs are large publicly held corporations. As being publicly held corporations they are required to publish a certain amount of information about revenue and markets to their customers. This information is publicly available since there is little point to keeping a secret, if that was possible. So the people that compile statistics can only extrapolate market shares by revenue shares.

But there are many major problems to this approach, like:

* It does not actually tell you what is being used for what. They may by a server, but you don't know for a fact that they are even using it for anything. As far as we know the majority of customers could be piling servers into a feild and setting them on fire. Now this is unlikely, but it would be impossible to really know one way or another.

* It does not tell you how long they are being used. People tend to swap out Windows servers 2-3 years. People tend to use Linux and Unix systems for much longer. But you can't know to what extent or how often that is the case with any sort of reasonable accuracy, and it's impossible to know much more beyond educated guesses.

* It does not reflect numbers of servers from sources other then purposely sold servers by major OEMs. So-called 'White Box' servers, which are popular, are sold by generally privately held corporations whose activities are not being tracked by research groups. Also you don't know about desktops-turned-servers, or people putting Windows on Linux servers or visa versa. And all sorts of things like that.

Personally I think that this means that Linux market is heavily understated. But it's impossible for me to know one way or another.

You may have noticed that Linux server revenue is closer to Unix revenue. However Linux servers tend to cost less then (guessing..) 5 grand while Unix systems can cost a half a million dollars sometimes, maybe even more. So the amount of Linux servers out there probably outnumber Unix systems 10 or even 100 to one.

When you compare Linux vs Windows server the licensing for Windows is such that going out and buying a dedicated server from Dell is cheaper the taking a older machine and upgrading it or buying a desktop and installing Windows server on it... which is all common things for Linux folks to do.

*shrug*

Unpaid deployments

Posted Dec 1, 2009 10:52 UTC (Tue) by robbe (guest, #16131) [Link] (1 responses)

Why not just *ask* a sufficiently large sample of businesses? Under
restrictions to only publish aggregate numbers, if needs be. You probably
won't get honest answers about their unpaid Windows deployments, but
there is no big incentive to lie about unpaid Linux installations.

I was under the impression that many (most?) of the Gartner/IDC/whatever
studies are done this way.

Unpaid deployments

Posted Dec 4, 2009 1:19 UTC (Fri) by dlang (guest, #313) [Link]

the people filling out the surveys are not always going to be the people who know what's running.

over the last 10 years it hasn't been unusual to see a company pronounce that linux is junk and then learn that a lot of their datacenter infrastructure had been moved to linux without the senior management knowing about it.

the people who know how many systems are running linux are too busy getting work done to fill out this sort of thing

Unpaid deployments

Posted Nov 16, 2009 16:31 UTC (Mon) by blitzkrieg3 (guest, #57873) [Link] (3 responses)

He said something along the lines of:

"If you asked me two years ago about the adoption of unpaid linux, I'd have said that it would mostly get converted to paid linux. That's not what happened."

The graph looked mostly linear for both. I could go into more detail but I believe he is going to post the slides somewhere.

Unpaid deployments

Posted Nov 16, 2009 17:23 UTC (Mon) by jspaleta (subscriber, #50639) [Link] (2 responses)

Just to clarify... which "he" are you referring to? Brian Stevens or Carlos Montero-Luque or were you referring to Al Gillen's comments during his IDC presentation?

-jef"pronouns kills"spaleta

Unpaid deployments

Posted Nov 16, 2009 17:25 UTC (Mon) by blitzkrieg3 (guest, #57873) [Link] (1 responses)

Al Gillen's comments during his IDC presentation

Unpaid deployments

Posted Nov 16, 2009 17:42 UTC (Mon) by jspaleta (subscriber, #50639) [Link]

That's an interesting quote about revising 2 year old market evolution expectations coming from the perspective of a 3rd party analyst. It would definitely be interesting to know how the IDC estimates the number of unpaid versus paid deployments for the purposes of producing the aforementioned graphs.

-jef