Making a Linux distribution is easy, and lots of people have done it and continue to do it. All you have to do is get the source code and integrate the pieces you like and slap your logo on it.
Making a commercial Linux distribution that makes enough money to cultivate innovation and stability in the kernel is not so easy, however. Very few companies have done it, but Red Hat is one of them – and its Red Hat Enterprise Linux (RHEL) marks its 10th birthday today.
RHEL, the commercial-grade operating system that the then young Red Hat created in the wake of going public on the dot-com boom, was not Linux for grown-ups. That's what the Fedora, OpenSUSE, Ubuntu, Debian, Gentoo and other development communities are for. No, RHEL was Linux for children.
By "children" I mean it was for corporations of the world that just want to install an operating system, plunk some apps on it, and manage it like they do other platforms – and be able to forget that it is even there because it just works.
RHEL emerged from the craze for Linux company IPOs that went hand-in-hand with the dot-com bubble of the late 1990s and early 2000s. Red Hat went public in August 1999 and Linux was on fire with Red Hat, as the first big Linux distro offering commercial support, the hottest.
Red Hat had lined up partnerships with Compaq, IBM, Oracle, Computer Associates, and others in the systems racket after Linux was enthusiastically embraced years earlier by academics and the supercomputing labs.
Making the Unix Wars irrelevant
Linux, being open source and out of anyone's control, made the Unix Wars irrelevant - some would say childish - and the world still wanted the promise of open systems (which was only partly accomplished). And it got open source to boot. Linux quickly became the system that ran on all iron – whether server makers or PC makers like it or not.
Red Hat Software was founded through the merger of Bob Young's ACC Linux, a software utility supplier for Linux and Unix tools founded in 1993, and Red Hat Linux, a popular Linux distro created by Marc Ewing, the original Shadowman (the nickname for the Red Hat logo) who roamed the halls at Carnegie Mellon University.
This was not a company you would have guessed would be valued at $19.7bn by Wall Street in the December following its initial public offering. But there was a reason why a company that only generated $10.8m in its fiscal 1999 year ended in February just exploded when it went public later that year.
After Red Hat came crashing back down to earth, and Wall Street with it, what was left was a company that was a credible threat to the Unix, Windows, and proprietary system incumbents that had an army of enthusiastic, idealistic, and sharp open-source coders who were not so much interested in getting rich as they were in being right and to have their peers concede that they were. (Well, that is what a meritocracy is, at least when it is working properly.) And you might not have guessed, at that time, that Red Hat would be the first company founded on open-source software to break through $1bn in annual sales and have $1.3bn in the bank.
Too bleeding edge
And it almost didn't happen. Matthew Szulik, Red Hat's president and chief operating officer and eventually chief executive to Young as chairman, led Red Hat to the IPO and through it, but the bleeding edge nature of the original Red Hat Linux was an anathema to a company trying to establish a commercial operating system business that was stable and safe in the eyes of the corporations that would shell out cash for that support.
The idea was to make Linux as good and familiar as Unix, which garnered the lion's share of system sales during the dot-com boom era (a little less than half of all systems sales worldwide were driven by Unix).
So in 2001, Szulik hired some expert techies with deep backgrounds at Digital Equipment Corp: Paul Cormier, who became executive vice president of engineering and is now president of products, and Brian Stevens, who was brought in to manage the Linux distro and who is now CTO at the company.
Among many other things that they have done, Cormier was part of the AltaVista search engine team – the thing we all used and loved before Google came along – and Stevens was one of the key developers for the X windowing system and Kerberos security and the architect in charge of Tru64 Unix at DEC. What they wanted to do was tame Red Hat Linux and turn it into Enterprise Linux.
"I don't think the general person really understands what this meant from 10 plus years ago," Cormier told El Reg, in an interview marking this month's 10th anniversary of RHEL.
Tough growth decision
Cormier: Making Linux more palatable and useful for corporations
"We were faced with a really tough decision back then, and we really bet the company on this and it paid off in a billion ways. We made a decision to stop Red Hat Linux, which at the time was the most popular, hobbyist Linux operating system out there. It was a very, very, very unpopular decision with the engineers and with some people in the community. The engineers asked the CEO to fire me at the time."
Over beers, the engineers named it Red Hat Enterprise Linux, and the idea behind it was to make Linux more palatable and useful for the corporations that ultimately pay the bills. At the time, says Cormier, Linux was only used by the most technically savvy companies, and only in their most technical departments.
"What we did to allow Linux to be consumed by the masses was combine the open-source model with the enterprise product model and while still keeping the open-source mantra," says Cormier. "The real science of the RHEL model is getting these backports into the kernel without breaking compatibility and stability."
And, by the way, this is how most operating systems work these days and why operating systems seem to be progressing at a snail's pace compared to a decade ago. Vendors and users alike don't want new versions and the new application certifications they require.
If Cormier had to do it all over again, RHEL might have come out concurrently with Fedora, the upstream development release of Linux that eventually became RHEL these days. But it didn't start out that way.
"I'll be perfectly honest with you," Cormier concedes, "Fedora was a reaction to RHEL to make sure we continued to service the community. RHEL gave the predictability and stability that enterprises really wanted, but the community was really concerned that they were going to lose something on the bleeding edge. The community rallied around that bleeding edge, and that in turn is what drives RHEL. RHEL came first, and then Fedora, and if we had been smarter, Fedora would have come first."
And it just goes to show you that you don't have to get it right the first time in business. But you do have to fix it when you get it wrong, and to Red Hat's credit, when it was at risk of alienating the Red Hat Linux community by developing RHEL, it backed up and started Fedora.
And in the end, this is a better model than either the community or the company could do alone, and one that fits both enterprise customers and hard-core developers.
Red Hat Linux Advanced Server came out back in May 2002 and was eventually rebranded to the name that the engineers picked, Enterprise Linux. And by 2003, the company had one million licences under contract on its Red Hat Network support service.
A decade later that number has grown to 2.5 million active servers under subscription, according to Cormier. That may not sound like a lot in a world with around 36 million servers in it, but it is hard to reckon how many servers are using Fedora as well as the CentOS, Oracle Linux, and Scientific Linux clones of RHEL. It is likely that the free versions of Fedora and RHEL are on at least another 2.5 million or so servers, and then these other distros like Canonical and SUSE Linux add in.
What it is safe to say is that as far as enterprise servers are concerned, Red Hat is the dominant Linux supplier by far. It is hard to say if Linux in general and Red Hat in particular drove the $9.6bn in servers configured with that operating system in 2011 according to IDC's numbers (up 18 per cent) or if the server sales were going to happen anyway and that Linux just got its fair 25 per cent of the revenues as an established alternative to Windows and Unix.
One billion part II
Whether Red Hat drove those Linux sales or not, it certainly is the main beneficiary of them, and that is what Wall Street cares about and why the company is back up to a $10.9bn market capitalisation again.
What is also clear is that with current trends, Linux-based server sales will match Unix server sales in the aggregate this year, with each having around $11.5bn in revenues, but Windows-based servers will drive more than twice that revenue. It is going to take a long time and some radical changes to have Linux catch up to Windows. Maybe Linux never will catch Windows, and maybe it just doesn't matter right now.
People make their operating system choices based on application software, budget, and skills; no one is interested in platform wars in a cloudy world that can and does support Windows, Linux, and sometimes Unix side by side on cloudy infrastructure.
As Red Hat has correctly identified, middleware, file systems, server hypervisors, cloud fabrics, and other aspects of the modern systems need to be RHEL-ified and Fedorated. This is what Red Hat is focused on, and by doing so, it might get to $2bn in sales in half the time the first $1bn took. We'll see. ®