Enterprise Linux: is it broken?

Posted Nov 4, 2004 3:10 UTC (Thu) by dang (guest, #310) [Link]

It is worth noting as well that redhat has consumed a decent amount of IP and devloped still more orf its own that it needs in order to market to enterprise customers, and it has returned a lot of that IP to the community as free/open software. I've never understood the redhat bashing, to be honest. Half the hype around open source/ free software is that it provides market opportunities; so redhat gets kicked for putting that model in play for *what* reason? Especially since redhat has consistently given assloads back to the linux community gratis.

Also worth noting that for enterprise customers, redhat's license amounts to chump change; and for enterprise sysadmins ( and CTOs, and sales reps, and customer support staff ), vendor support is the difference between a painful incident report and an unfortunate pink slip ( or blown account on the sales or support side ).

Posted Nov 4, 2004 4:06 UTC (Thu) by hmsjohnd (guest, #24241) [Link]

At my company, we were frustrated for a long time, and hesitated moving from 7.3 to RHEL. At the end of the day, our decision was based on business. We think of Red Hat as an MSP; they provide updates, stability, and compatibility with 3rd party vendors. We focus on our core business model, and we can still make necessary adds and changes to the kernel, OS, and packages to suit our needs. RHEL also supports releases for longer than the hardware is going to be assigned to one function. When we have servers that have literally been up for years, this means a lot to us!

I wouldn't recommend using RHEL without the Satellite, because it's one of the major reasons we chose it. If someone wants to make a free Linux distro better, think about adding a full-featured web portal like that. Provisioning... PXE/Kickstart... Sure, we can go out and do that on our own, but will our hodgepodge of open-source tools be tighter-knit? Doubt it. RHEL is successful because it's practical - perhaps some of the free distros need to look at the big picture. Red Hat isn't *taking* anything from the community, it's providing an example of what customers are willing to pay for. How about a free alternative? If your distro du jour comes out every 6 months and is either too cutting edge or too conservative, why are you still wondering about RHEL's success? Where is the free distro with Enterprise bundle? I'm not talking about bloat, I'm talking about intrinsic, modern "Enterprise" functionality.

I don't think people look at TCO and ROI. Dog and Pony show aside, RHEL probably costs you less money over time than hiring employee(s) to duplicate the functions.

The Red Hat reps will negotiate pricing. Obviously, there are volume discounts. If you never really use support, and you have a central server for updates (no bandwidth leeching), are you really costing them that much resource-wise? No. Start off small, download updates manually... No legions of lawyers with red fedoras or BSA agents are going to show up at your office.

For the record, almost all of our admins are die-hard Slackware, Debian, Gentoo, etc. users on our laptops and on multiple home machines. Seeing a "dialed" network serving tens of gigabits of traffic with hundreds of machines usually converts non-believers, *at least at work* ;)

Posted Nov 4, 2004 4:36 UTC (Thu) by hp (subscriber, #5220) [Link]

Related web log entry: http://log.ometer.com/2004-09.html#5

Posted Nov 4, 2004 4:46 UTC (Thu) by csamuel (✭ supporter ✭, #2624) [Link]

> The per-CPU nature of RHEL subscriptions irks some people in the
> community. The restriction applies to support, however.

I'm not sure if that's completely correct, it also includes a license
for the Redhat trademarked logos and *I believe* (though it is very hard
to verify this) that this is per node, and if you install it on more than
the number of systems you have purchased for then you are breaking the
trademark license and can be sued.

However, Redhat do say that you are completely free to alter the SRPMS to
replace the trademarked logos and build RPMS from those, and then you are
free of these restrictions.

Hence how the NPACI Rocks cluster distribution is able to continue by
rebadging everything (as does White Box Linux I believe).

> If you just want the security updates, just get them directly from Red
> Hat's advisories and install them yourself.

Again I believe you have to be careful if the update contains trademarked
content.

> Red Hat has imposed no restrictions on the software which are
> inconsistent with its licensing; it is hard to see who is harmed
> by its activities.

The software continues to be free, and you are able to rebuild the RPMS as
long as you strip the trademarked logos, and even RMS doesn't seem to have
a problem with this.

Redhat have done a lot for the free software community, including having
purchased software and open sourcing it (viz GFS, and hopefully soon with
Netscapes LDAP server), and to do this sort of thing they need to make
money.

The main issue I've seen is with clusters, where paying per node per year
can really kill your budget.

It's really interesting to see that the next Ubuntu release will include
Kickstart support.. ;-)

NB: I reserve the right to be wrong on some or all of the above.
Corrections welcome!

Posted Nov 4, 2004 6:58 UTC (Thu) by ninjaz (guest, #2083) [Link]

Personally, I think the plan in Lineox's whitepaper sounds reasonable. The bootstrapping strategy (base it on a corporate standard - Red Hat) is basically the same method Mandrake used to get started.

Regarding drawing resources away from RHEL, I believe RHEL's core market will continue to use RHEL and pay its subscription fees as a matter of religion. Lineox looks like a different take on UserLinux's general idea, but with a migration path & compatibility and without the divisive 'one true way' baggage in each software category (with regard to GNOME/KDE, perl/python, etc) or Debian (and their reluctance to make releases with problems supporting newer hardware as a result... and just plain not being Red Hat, which some vendors hold a prequisite for getting support)

As stated in the whitepaper, Lineox plans to give back as well. But, for customers, provide an attractive choice without giving up on the chance of using proprietary vendor software (which experience suggests a manager type will choose sooner rather than later)

I think competition for vendor support would be a good thing. My experience with 'enterprise' distribution vendor support leads me to believe that the support is primarily a security blanket. The actual service levels have been well below what can be found on mailing lists.

A company strongly tilted toward providing support on a standard distribution could fill the need for organizations which don't have the luxury of expensive security blankets. It could even go the path of providing supplemental OS support for Red Hat customers in cases where Red Hat support has decided a particular support request is not worthy, but the purchasing committee has set Red Hat and their support as a baseline requirement.

What I find especially encouraging is the practicality of the whitepaper. Instead of beginning with a utopian dream or a huge ad campaign with skimpily-clad models (ala LinuxCare) the plan presented addresses the sort of issues I've seen in my experience as sysadmin.

Regarding the arguments against vendor lock-in, I don't find it surprising that they continue. After all, that has been one of the Linux scene's refrains all along.

Posted Nov 4, 2004 7:31 UTC (Thu) by BrucePerens (guest, #2510) [Link]

The Lineox white paper gives me credit, in small type. It's the UserLinux paper ported to their company. But the UserLinux strategy is not to bind support to one company, but to many in an open market, which addresses the problem in a way that the Lineox proposal does not.

Bruce

Posted Nov 4, 2004 14:16 UTC (Thu) by madscientist (subscriber, #16861) [Link]

I do agree with other posters that Red Hat has done a huge amount to advance Linux kernel development, as well as lots of other OSS and the idea of a Linux-based desktop as a whole. Not only that but they _do_ serve as an example that it is possible to make money with OSS, and that's very important. I think they've done a very good job and I don't begrudge them their fees.

However, there are problems... I actually don't think the problems are so much the fault of Red Hat per se but they are the central figure. The problems are not with the amount of money being charged; that's missing the point. The problem, as Bruce Perens so accurately stated, is with VENDOR LOCK-IN. If you are an enterprise customer and you want to use a Linux-based server or desktop you really have an almost impossible uphill battle to use anything but Red Hat. All enterprise customers have 3rd party applications that they want to run on their servers and desktops: things like ClearCase, Oracle, etc. These 3rd party applications are currently certified and supported only on specific versions of Linux. While most 3rd party applications do support more than one Linux distribution, the only one you can be 100% guaranteed that everyone, all the time, will support is Red Hat. This is a very difficult barrier to overcome when dealing with enterprise IT organizations. This gives Red Hat a huge amount of clout in the Linux space: it's hard to select anything else and once you're there, it's hard to move away. I think the idea behind LSB, UserLinux, etc. is exactly correct: what we need is a credible way for 3rd party applications to be certified on a platform which is not tied to a specific vendor. This allows a true choice when it comes to support options.

I do have to take exception to some of the responses regarding Red Hat though: we are a pretty large enterprise customer of Red Hat, with hundreds of RH servers and thousands of desktop deployments. I must say that our experience has not been all peaches and cream. The idea that you can make "changes to the kernel, OS, and packages" is just wrong: if you modify these things, especially the kernel, then you have voided your support contract with Red Hat and they will not support you. One example: we have a requirement to run UML on our desktops and under no circumstances will Red Hat support the inclusion of the SKAS patch into our kernels (for Red Hat Enterprise Workstation 3), even though this patch is fairly small and straightforward. We face the choice of either changing our entire simulation model, or going outside of support. Given the cost of support, which is not at all cheap, this really chafes. _IF_ we had an alternative for support then Red Hat would be forced to be more competitive and perhaps our experience would be better.

A single (for practical purposes) vendor for Linux has all the same issues as a single vendor for anything else.

Posted Nov 4, 2004 15:35 UTC (Thu) by cdmiller (subscriber, #2813) [Link]

I have no problem with RedHat's strategy. It is a side effect of becoming a public company. The focus shifts a bit more towards profits and the bottom line. Hopefully it won't distract from producing great free software. I suspect if their profit motive does eventually distract from keeping up the software, some developers will find other places they would rather work.

There is no such thing as vendor lock in with libre software. We had no trouble moving our servers from RedHat 9 to Mandrake 10 when we found RHEL priced out of our acceptable range. We also flirted with SUSE and debian. Several 3rd parties provide the same levels of support for Mandrake etc. as they do for RedHat and Fedora.

Posted Nov 5, 2004 1:25 UTC (Fri) by garloff (subscriber, #319) [Link]

It's a bit funny to see arguments telling that Enterprise Linux is
unneeded and that a free distribution with some support from a 3rd
party would provide the same to the enterprise.

Let me give you an idea what's happening before an Enterprise Product
is released:
* integrating all components into an installable and managable
solution
* 6 month beta-testing and stabilization period
* heavy functional testing and stress testing
* ensuring with IHVs that server hardware is supported and functioning
well
* working with ISVs and ensuring their products work
* certification testing
* pushing some of the features that's wanted by many partners
* backporting a few features that have been implemented upstream
meanwhile and that are safe to backport

This obviously includes heavy engineering work to get all problems
resolved and fixed. A company that produces an enterprise Linux and that
I happen to know very well did create a couple of hundred bugfixes for
the Linux kernel during that period; almost all of which ended up being
merged upstream.

When the product gets released, there's HW and SW certified for/on it,
which gives some assurance to the user that the products will work and
that he can get support for it.

After the enterprise Linux is released, there will be
* Maintenance: Bugs and Security Problems get fixed; service packs
may add much-asked features if they don't put the stability and
interoperability with 3rd party SW at risk. And this is provided
for many years.
* Support: The engineers that have fixed all the bugs during the release
cycle know the code and can efficiently help with tracking down critical
defects, so providing good bug resolution times.

I seriously doubt that the same level of quality, stability and support
can be provided by a company offering support for e.g. Fedora Core or
Debian Linux.

I can see why the community does not need Enterprise Linux. The community
consists of developers who need cutting edge technology rather than
ultra-high stability and support offerings with short term response times.
I definitely don't need enterprise Linux on my machine I use for
developing the kernel or Python apps.

But I see no reason why one shouldn't acknowledge that there's businesses
out there that need Enterprise Linux. There obviously is a market. RHEL is
doing well there and there's at least one more company that does well
there also.

I rather see some slightly obscured trial to get part of the enterprise
business without needing to put the same level of effort into
engineering, stabilization, testing, certification, maintenance and
support than the current Enterprise Linux offerings do. Nice try.

Posted Nov 11, 2004 7:25 UTC (Thu) by dbunker (guest, #267) [Link]

Though I find RedHat's use of trademarks to enforce their policy wishes to be somewhat shady, I don't question that they have that right. I am saddened, however, that RedHat has essentially forced me away from using their distribution.

I work for a semi-large company with several hundred employees, but we don't have tens of thousands of dollars to throw around; every year we have to fight a war to keep our IT budget where it is. Over the last seven years, we've installed various versions of RedHat on several dozen servers and have been very pleased with it. Now we're not sure what to do -- be don't have a lot of money to buy licenses, and we don't want to reinstall a new Fedora Core every year and a half. We could have managed to buy something at a reasonable cost (i.e. a one-time cost of $300 for each server, or maybe a yearly cost of $30-50 per server), but unfortunately, RedHat does not seem to want to cater to that price range. They seem to think that there are only two kinds of "customers": rich and poor. There is an entire world in-between that they just don't seem to care about -- those people who don't care about Oracle certification or cluster file systems, but rather just want some refinement and critical security updates for three years or so.

I complain about RedHat's new strategy not because I think they're cheating me -- I can find something else to use -- but rather because they have decided to ignore my business. I've enjoyed using RedHat's products and I would love to purchase an operating system from them, but it just doesn't look like a possibility now. I just don't understand why they won't offer something between Fedora Core and RHEL ES.