Google considered a move to OpenSolaris? All that glitters is not Chrome

CORRECTION at 6:30 a.m. PDT September 2: This blog inadvertently had linked to old information about Google's work with OpenSolaris. However, having discovered the mistake, the author realizes how salient the content is to Google's announcement Monday of its new browser.

Two years ago, Computerworld reported that Google was actively testing Sun Microsystems' OpenSolaris Unix distribution as a possible adjunct operating system to be used internally with its existing modified Linux distribution. While I'm sure there continues to be active experimentation at Google around OpenSolaris, I suspect any move away from Linux remains highly unlikely, at least in the short term.

In a similar vein, Monday's news of Google's creation of a new Web browser--Chrome--may not spell the end of Google's cozy relationship with the Firefox browser either.

According to the 2006 Computerworld article:

Sources outside Google of said that its servers currently run a stripped-down version of Red Hat Linux that has been modified by the company's engineers. A Solaris systems administrator who recently interviewed for a job at Google said that he was told by employees there that the search engine vendor plans to create and test its own modified version of OpenSolaris....

Switching to OpenSolaris would be a natural move for Google, which has a large number of former Sun employees and is striving to push the performance of its data centers, (technology consultant Stephen) Arnold said. But he added that he doubts Google is widely deploying OpenSolaris yet. "Will it quickly replace Linux anytime soon? No," he said.

Exactly. Google is the performance king, and so it might have been willing to make a bet on OpenSolaris that others (like eBay and Yahoo) also made. Solaris has long been considered the gold standard for performance.

But two years later, Google has yet to broadly embrace OpenSolaris. Google isn't one to take the short-term view on performance. Linux has a strong, vibrant community dedicated to improving its performance and extending its reach. OpenSolaris, while a great project, still lacks this widespread community involvement. In Linux, Google benefits from the contributions of Hewlett-Packard, IBM, Red Hat, and others. With OpenSolaris? It would be more of a solo act.

This is just one reason I think that Chrome is unlikely to displace Firefox in Google's affections, at least anytime soon. OpenSolaris (and Chrome) may have technical superiority to offer them, but they have nothing to offer in terms of market momentum. And Google knows how to read the momentum tea leaves.

[bdwarner]

Hi Matt,
Just a friendly pointer - that article on OpenSolaris is from 2006.
Best,
Brian

[fazalmajid]

OpenSolaris' chief advantage over Linux is not so much speed as robustness, specially under load, and observability. Google has hundreds of thousands of servers, possibly millions. The ability to run them at higher loads without buckling catastrophically under pressure means fewer servers are required, and thus capital savings and more importantly electrical power and HVAC savings.

Dtrace (also available on FreeBSD and OS X but not Linux due to the latter's not-invented-here mentality) would also give Google the ability to extract fine-grained performance stats from its apps to optimize them further.

As for the "many eyeballs" fallacy, you're better off with a tight band of highly skilled coders than a crowd of middling ones. Solaris was by far the best of the proprietary UNIX variants, vs. HP-UX, AIX or Tru64. It was also the only one available on a x86 architecture and one of the first OS optimized for AMD64 and NUMA due to Sun's heavy use of Opteron. For all these reasons, Sun's implementation on industry-standard architectures (as opposed to oddities like zSysytem mainframes) is far more robust than what HP or IBM can bring to the table.

[odubtaig]

Of course, license incompatiblity which is requiring a substantial rewrite of all kernel-level code on both DTrace and ZFS before they can be used isn't at all a barrier. Nope, it's all egos and NIH.

I especially love when people claim ZFS is the victim of NIH syndrome completely ignoring the widespread support for ReiserFS and XFS which were invented by one man in his bedroom and SGI.