June 17, 2004 | By the time you read this column, RedHat will have officially killed the last free version of RedHat Linux. The death of RedHat 9.0 on May 1, 2004, leaves the company free to focus on Enterprise Linux, a product line that requires hefty per-machine fees each and every year.

In trying to accommodate these changes, the research computing community is dealing with two significant issues. The first problem is a rigid pricing structure that places an extremely high price premium for the use of RedHat Enterprise Linux on anything but low-end 32-bit machines with Intel processors. The tiered pricing as it exists today seems biased toward corporate clients deploying powerful enterprise application or database servers. For these clients, a few thousand dollars in extra OS licensing costs per year is trivial when compared to the cost of licensing and deploying the application software itself. Research organizations, however, commonly use large numbers of Linux servers with hardware attributes (64-bit AMD or Itanium2 architectures, 4-8 CPUs, very large memory) that vault them into the Enterprise Linux pricing tiers requiring thousands of dollars per year, per machine, in new licensing fees.

The new pricing structure turns what used to be an easy research IT decision into one requiring a careful cost-benefit analysis.

The second major problem is the forced bundling of Web- and phone-based technical support with access to product updates and security patches. Those of us involved in research or technical computing tend to use and customize Linux in ways that pretty much guarantee a scripted response of "I'm sorry, we can't support that" whenever we are dumb enough to call the Linux OS vendor support line in the first place. Knowing this, we tend to support basic OS installation, configuration, and troubleshooting issues internally while relying on the vendor to provide software updates and security fixes.

With the Enterprise Linux product line, there is no way to purchase the product (and update services) without the bundled OS support. The limited 30-day support option in the "Basic Edition" looks attractive at first glance but applies to only a small portion of the product line.

The all-or-nothing bundling of OS support with access to critical updates and patches combined with expensive annual fees is forcing me and my colleagues to seek non-RedHat alternatives wherever possible. Organizations that were previously 100-percent RedHat Linux-based are now forced to find a new standard Linux platform while rationing out Enterprise Linux licenses only where business or commercial software requirements demand it.

Fortunately, RedHat is making this change in 2004, when it is still fairly easy to migrate to far less expensive alternatives. Had it made this move a few years later, the transition would be much more painful.

At the heart of this struggle is the fact that RedHat Enterprise Linux is a great product, especially with the potent combination of a 12- to 18-month major release cycle backed up by a five-year product lifetime. I've fallen off the constant OS upgrade treadmill so many times now that I'd happily jump at the chance to pay for a product with a five-year life span and 12 to 18 months between major updates. RedHat's extensive hardware and software process also provides real value to researchers needing certified systems on which to deploy large databases, storage infrastructures, and complex commercial software. The new pricing structure, however, turns what used to be an easy research IT decision into one requiring a careful cost-benefit analysis.

So what are the alternatives — particularly for organizations with existing 32-bit systems and plans to bring in more 64-bit horsepower? For those who wish to stick with existing RedHat 7.1-9.0 systems, a company called Progeny will provide updates and security patches for $5 per month per machine. The "transition service" from Progeny can provide some much-needed breathing room for groups still figuring out internal Linux strategies. The Fedora Legacy Project is also trying to provide updates for a few orphaned versions of RedHat Linux.

For the past several months, we at The BioTeam have been evaluating RedHat alternatives in our hardware lab. The two distributions that most impress us so far on both 32- and 64-bit systems are Suse and Tao Linux. Purchased recently by Novell, Suse is a fantastic product, supported by many commercial software vendors. Suse 9.0 deployed and ran great on all the hardware platforms we tried. AutoYaST in particular is a breath of fresh air for those used to hacking RedHat Kickstart files to customize automatic server deployment and configuration. It may become our new Linux OS of choice for future projects.

We have been evaluating RedHat alternatives in our lab. The two that most impress us are Suse and Tao Linux.

The other interesting find was Tao Linux, one of the emerging crop of "generic" Linux distributions that cannot legally be called "RedHat" but are all built from source RPMS legally downloaded from the RedHat Enterprise Linux repository. It is interesting to think of how Tao and the other generics could be mixed and matched where appropriate with officially purchased RedHat Enterprise Linux systems. This could keep Linux cluster costs reasonable, as the critical machines run the official Enterprise Linux AS product while the compute nodes all run Tao Linux variants.

I suspect that we will continue to use RedHat Enterprise Linux in situations where commercial software requirements force the issue, and perhaps also when deploying systems that have to connect to complex storage or network infrastructures, as using a certified OS can simplify the finger-pointing loop when things go awry. Personally, I'm also still recommending RedHat to at least two different groups of people: those in research institutions who have managed to negotiate reasonable site license agreements, and occasionally to large organizations that can gain enough ROI from the slow release cycle and five-year product life span to justify the high annual fees.

What are you going to do?

Chris Dagdigian is a self-described infrastructure geek currently employed by The BioTeam. E-mail: chris@bioteam.net. 


ILLUSTRATION BY MICHAEL KLEIN