TurboLinux announces lay-offs and refocus
But coming late to the Linux cluster party could cost
There was some barely disguised meerschaum in some quarters of the Valley after TurboLinux announced lay-offs: reportedly ten per cent of the company's workforce.
The layoffs "were significant enough to affect profitability and shallow enough that we can focus on what we do best", according to TurboLinux president Paul Thomas, who identified these as clustering and Japanese language versions of Linux, according to G2.
That sounds a 'bad-news, worse-news' gag. Although the company is targeting niche markets, those niches are the same ones that were long ago identified by Red Hat and SuSE, who've been investing and recruiting accordingly.
TurboLinux received its first outside funding a couple of years ago, back when it was Pacific Hi-Tech, barely out of dabbling with wallpaper CDs and games (ah, like the Elephant, the Vulture never forgets...) but received a round of blue-chip funding at the height of Wall Street's infatuation with Linux stocks.
So why the uncharitable comments and ill-will? Much of that is generated by TurboLinux's take on the open source model, which has involved a combination of time-delay source release and its own proprietary features. Despite the outrage, the latter is actually pretty harmless - for example management GUIs are released only as binaries, not as source - and gives the company some time to market advantage but doesn't really preclude smarter alternatives. But the rather grudging time-delay ploy has led to several high profile accusations that TurboLinux is playing by the rules rather than spirit of the game. That developers don't like it is one thing, but open source developers tend to need code fast, with the result that TurboCluster Server has become increasingly marginalised.
And moreover, it's far from unique. It's maintained that the company took published open source IPVS code written by a Hong Kong student Wensong Zhang, reverse engineered it, and called it their own. Zhang now leads the Linux Virtual Server project and is a consultant to the Red Hat Kernel Development Team.
(And it was the release of this patch that got some outsiders into a lather about kernel-forking last year, based on the unlikely premise that Linus would include TurboLinux's ipvs into the kernel. Which in fairness TurboLinux was quick to deny.)
TurboLinux seems to acknowledge that it's come late to the open source cluster party. The real action over the past year has been away from the limelight, in devising a VAXish, shared-everything cluster model suitable for high-availability databases and transaction processing. Which is a lot harder than load balancing a bunch of web servers.
So following his landmark speech at the New York LinuxWorld earlier this year, TurboLinux hired file system giant Peter Braam - one of the two key architects behind the next generation clusters, and a member of the original 'Cluster Cabal' - to work in its New Mexico office. We can't for a moment imagine that Braam, whose own track record - from his Carnegie Mellon days (leading the Coda file system work) to his own Intermezzo file system work - is quite impeccable, tolerating wheezes such as time-delay release, so that's a good sign that TurboLinux is cleaning up its act.
But the company needs corporate mindshare for this kind of data centre revenue, and for now it looks like SuSE and Red Hat are far better placed to sell into these accounts than TurboLinux. So reluctantly we have to conclude that to stay in the big league of Linux distros, convincing webmasters and ISPs of your merits isn't quite going to be enough any more. ®