Sun: OpenSolaris 'pretty freaking amazing'
On the road from Indiana to Nevada
While the top brass at Sun Microsystems might seem to be clueless about the company sometimes, there are plenty of people who know where the bread gets buttered - or doesn't. They know that Solaris and the servers that run it are what really matter at Sun. And that means Sun's OpenSolaris project and its related Solaris commercial distribution are still the key to success or failure for Sun Microsystems.
The best example I can give for Sun's silliness is the company's decision last August to ditch its SUNW symbol, which showed Sun's heritage as a workstation maker, on the NASDAQ stock exchange. Instead of picking the real symbol of the company, which you could abbreviate SLRS, Sun's president and CEO, Jonathan Schwartz, decided to make his mark (well, one of many) on the company by switching it to JAVA. Like Java ever really made any money for Sun. Lots of good PR since 1995, no doubt. But money? No way.
When we look back a few years from now, we might see that Sun's decision to open source Solaris in early 2005, when Solaris 10 came to market, was a turning point for the company. Not because open sourcing of software fixes a lot of problems, but because Sun stopped trying to use its leverage on legacy applications running on Solaris to peddle inferior Sparc iron and returned to a strategy of embracing x86 (well, x64 by then) servers and letting Solaris live wherever it could.
At the same time, Sun put better entry and midrange Sparc iron into the field, was working to get better high-end iron out with partner Fujitsu, and got to work getting a competitive x64 server portfolio to market. The open sourcing of Solaris through the OpenSolaris project gave Sun a change in momentum, and it gave customers the confidence that Sun was serious about changing and, more importantly, that Solaris could survive even if Sun did not.
Sun is, of course, still around. And the OpenSolaris project is bringing in lots of community members, while at the same time Sun and its collaborators are working to get the next release of Solaris out the door - perhaps by next year. For those of you who don't know, Sun is backing its way into a Linux development model, with a development release and a commercial release, akin to the Fedora and Enterprise Linux or openSUSE and SUSE Enterprise Linux Server splits over at Red Hat and Novell.
150,000 developers strong?
Recently, Sun was touting the fact that more than 150,000 people have joined the OpenSolaris community. This is a pretty big number, particularly since the first release of OpenSolaris, dubbed 2008.05 and developed under the code-name "Project Indiana," only started shipping in May. That number doesn't tell the whole story.
For one thing, it is not a measure of downloads. According to Dan Roberts, director of marketing for Solaris at Sun, the project allows completely anonymous downloads of OpenSolaris and also uses mirrors to distribute the code. OpenSolaris 2008.05 accounted for a couple hundred thousand downloads from the project's systems and maybe half again as many downloads on mirror sites.
"That's pretty freaking amazing for a large-scale operating system," says Roberts, adding "that's well above what we did for Solaris 10 in the first month." Specifically, it is about three times the initial download rate for Solaris 10 back in early 2005. Given a slowdown in downloads that happens every summer, Roberts estimates that a high number of multiple hundreds of thousands of OpenSolaris 2008.05 downloads have been distributed by the project and its mirrors to date, but that figure is probably short of the 1 million mark.
What's more, that 150,000 developer number doesn't tell how many people are really contributing to code day in and day out, so it's not a good number to just throw out there - unless you like to be vague. (As IT vendors often like to be in their press releases).
Let's drill down a bit. Sun has a little more than 1,000 of its own software engineers working on OpenSolaris, which includes the coders hammering away on the Solaris kernel, networking stacks, file systems, and such. This number does not include application stacks that ride atop Solaris or other operating systems, or Sun's compiler teams.
Outside of Sun, Roberts says that there are a few hundred outsiders, including techies from other IT companies (think Intel and AMD and such) who contribute. This is the core development team for Solaris. Now, outside of that, there are several thousand other community members who are very active, working on documentation, internationalization, or sub-communities like storage. And the rest are just the OpenSolaris masses, who file bug fix reports and answer questions on the forums to help fellow users out.
Sun is very happy with these numbers, by the way. "When you build an open source community, you don't plan to have 100,000 people submitting code," Roberts says with a laugh. "And if you did, you would probably want to kill yourself because you wouldn't be able to get anything done."
Time for a merger
Sun definitely wants to get things done and is working toward the day when it can merge the OpenSolaris project with its internal "Nevada" development release (what we'll now call Solaris 11). Back in early 2005, when Sun launched Solaris 10, it took the code base and started making changes for Solaris 11, developing it internally as before, while at the same time open sourcing Solaris through the OpenSolaris project. To make OpenSolaris the true development foundation for Solaris, the internal and open development efforts - and their code - need to be merged.
Sun is not saying when this will happen, of course. But what Roberts will say is that OpenSolaris 2008.11 is on track for November this year, and that subsequent to that, Sun will shift to an April and then October roll-out schedule for OpenSolaris releases. This will get Sun better in phase with school and vacation schedules around the world.
He also said that Sun will sometime next year take all of the goodies in Nevada and OpenSolaris and merge them in the wake of one of the releases for OpenSolaris. That doesn't mean that Solaris 11 will be launched at that time, but rather that there will be a single Solaris development release that can become Solaris 11 when it goes commercial.
At that point, OpenSolaris will be on a six-month release schedule with commercial support, while Solaris proper will be akin to the Long Term Support releases done by Canonical with its Ubuntu variant of Linux. Except that Sun will be providing a minimum of 10 years of support for Solaris, rather than the 5 years that servers get on the Ubuntu LTS releases. In some cases, Sun has supported a Solaris release for 11 or 12 years - something that its customers often require, given the long-term nature of the servers they install.
Sun is not saying much yet about what OpenSolaris 2008.11 has in store, but Roberts raised the curtain a little. The Web-based image packing system that was part of the initial OpenSolaris release from May was pretty basic, and the next release will have a script-driven, network install feature that will make it easier for system admins to use OpenSolaris on their networks of servers and do snap upgrades on those instances.
OpenSolaris 2008.11 will also have the first phase of the distribution configurator, a custom OpenSolaris packaging tool like those available for Linux that allows users to create their own distros. Sun will also tweak the repository code in its OpenSolaris build system so users can store their custom distros locally on their own networks while still using that build system. And finally, there are some elements of Solaris that cannot be open sourced (because the companies that own the code are dead and gone) or won't be (such as video drivers or fonts) that Sun nonetheless wants OpenSolaris shops to be able to use. So these elements will be distributed from the Sun site, which has licenses to do so in binary form. ®