Solaris Nowhere Close To Death

By Chris Josephes
September 30, 2008 | Comments: 4

If you wanted to obtain an unbiased viewpoint on Solaris, would your first choice be the Executive Director of the Linux Foundation? He does offer a viewpoint, but it's probably not unbiased.

Let's get the obvious part out of the way. Zemlin's job with the Linux Foundation is almost the equivalent of Steve Ballmer's job with Microsoft. Of course, the humorous part is that Jim is doing his job in the style of Steve Ballmer. That includes loud, outrageous, and sometimes incorrect claims that are easy to refute. They're both using the press to rally their user base. And if they can't brag about the merits of their own product, they point out the faults of their competitor's.

First, Zemlin drives a huge wedge between Linux and Unix, claiming Unix (I assume this could mean AIX and HP-UX as well) is a legacy architecture. That statement can be a little muddled, because a technical person would wonder whether he's talking about the user-land space or the OS kernel. The point he's driving at is that Unix is old, and old is bad.

However, it would be fair to point out that the Linux kernel is 17 years old as well. Also, age of code has nothing to do with quality of code; nor should anyone suggest that all code depreciates equally. There is good code out there today on many platforms that stills work the same today as it did 20 years ago.

Next, he claims that features like ZFS and Dtrace are minor features that nobody wants or cares for. That would be easy to refute with outside sources, but Zemlin does a good job of refuting the claims himself. Zemlin believes that both of these unused features should be put under the GPL license so they could be integrated into Linux.

If ZFS and Dtrace are features that nobody wants, why would the Linux Foundation want to include them?

Maybe it's because SystemTap is not turning out to be as impressive as they hoped. Google shows 800,000 entries for Dtrace, compared with 165,000 for SystemTap. Dtrace has already proven itself in the field by improving the performance of large scale web applications like Twitter. Hundreds of end users are contributing dtrace scripts to make it even more useful. SystemTap is going to need more time to mature before it gets widespread use in the user community.

As far as filesystems go, Linux is a little behind. It's pretty clear that Ext4 will probably be the next big thing in Linux, but it's not production ready. Ext4 still doesn't scale to the sizes offered by ZFS, nor does it include the management and recovery features.

Zemlin is concerned that these tools provided by Sun are gaining in popularity both inside and outside of Solaris. Once Dtrace and ZFS start building traction in FreeBSD and OS X, they will benefit from more contributions and community building. And the best way to stop that progress is to dismiss the community entirely. Pretend that they don't exist and that their code doesn't even matter.

Which is another trick that Zemlin probably learned from Ballmer.

You might also be interested in:


Zemlin's obviously trading on the fact that Linux is percieved to be young and vigorous compared to Unix, where for practical purposes there's as little difference between Solaris and Linux as there is between Solaris and HP-UX or AIX.

As a long-time Solaris admin coming to Linux a few years ago, it is massively more user friendly and easy to configure than Solaris which by contrast has a clunky terminal and outmoded standard components such as package management, SNMP and UFS.

We recently had a power outage in our data centre. All the Lunux boxes booted up fine, but half the Solaris boxes had crippled boot filesystems in need to fsck-ing. It's all very well having super-duper ZFS arrays, but if your boot filesystem is a creaking relic then you've got problems.

Youre right, Solaris has some great features, but many areas are in severe need of an update.

I agree with Simon Hibbs comments. We ran Solaris for about 5 years along with our Linux boxes, and it was always the Solaris ones that were the hardest to deal with either with them crashing, recovering from a crash, or even training new sys admins on. Over time we started questioning ourselves about why we even bothered to run Solaris, and have now finished migrating to a full Linux environment.

Sun has some great engineers with tremendous ideas, but they lack in the implementation and design parts of their solutions. It shows in Solaris as well as the rest of their enterprise applications.

Pre-Solaris 10, i agree with all previous comments.

Solaris Express (and Opensolaris) both support ZFS root now.
But straight Solaris has had journalled UFS journalling enabled by default for as long as I've used it, so the fsck comment makes me think we're comparing Linux to very old Solarises.

Projects like coolstack give you an awful lot of common services (Rails, memcached, squid, a LAMP stack) out of the box, I don't see any real need to be bothering with Linux anymore.

First of all, I am an employee of Red Hat, but whatever I write here is just my opinion and I don't work anywhere close to anything which would make me able to do comments in any manner resembling anything official. These are just my comments and nobody else.

Having said that, I have to point out one difference between Jim Zemlin and Steve Balmer. When Steve speaks, he really speaks in the name of Microsoft. When Jim speaks nobody cares because he does speak only in the name of himself and suspicious legal entity which was created for different purpose than to be leaders of the free software community (i. e. to be the employer of Linus). I am very gratious that free software (or Linux, if you wish) community is not lead by him.

News Topics

Recommended for You

Got a Question?