[LWN Logo]

Bringing you the latest news from the Linux World.
Dedicated to keeping Linux users up-to-date, with concise news for all interests

 Main page
 Linux in the news
 Linux History
All in one big page

Other LWN stuff:
 Daily Updates
 Linux Stocks Page
 Book reviews
 Penguin Gallery

 Use LWN headlines
 Advertise here
 Contact us

Recent features:
- RMS Interview
- 2001 Timeline
- O'Reilly Open Source Conference
- OLS 2001
- GaŽl Duval
- Kernel Summit
- Singapore Linux Conference
- djbdns

Here is the permanent site for this page.

See also: last week's LWN.

Leading items and editorials

Cray goes into the Linux cluster business. Cray has long been the definitive manufacturer of "big iron" supercomputer systems. From the original Cray 1, released over twenty years ago, through to upcoming products like the SV2, Cray has been the king of huge, expensive, blindingly fast systems - though competition from Japanese supercomputer makers has tarnished that reputation somewhat in the last decade. Your author, who had the dubious honor of programming on Cray 1 serial 3 (using punched cards!) many years ago, can get very nostalgic about Cray's heyday.

The big iron market is a hard place to be, however. The technology to produce that kind of processor is extremely expensive to develop and manufacture; note, for example, that the development of the SV2 is being supported by the U.S. government, and has been underway for quite some time.

But the real problem, of course, is that clusters made from cheap, off-the-shelf components can achieve supercomputer performance for a fraction of the cost. Many one-time big iron customers are realizing that [Cluster] they can save a lot of money by installing clusters. Some problems still require big iron, but many do not; thus we see companies like IBM and HPTi winning high-profile supercomputing contracts with cluster systems, and Cray often being left out in the cold.

No longer. On January 29, Cray announced that it will soon begin selling Alpha-based Linux clusters. The company has not yet released a whole lot of details on the "SuperCluster(TM)" product, but a few things have come out. It will be built using dual-processor, 1U Alpha-based CS20 servers from API Networks. A Myricom interconnect will be used, and Cray plans to scale the systems up to "thousands of nodes." One customer (BioNumerik Pharmaceuticals) for the new cluster systems has been booked already, with others expected by the time the systems ship (toward the middle of the year). The SuperCluster is, according to Cray's Steve Conway, the successor to the company's popular T3E parallel Alpha supercomputer; the average price for one of these clusters is expected to be "in the single-digit millions."

On the software side, Cray will be adding management software for tasks like failure rollover, job checkpointing, and accounting. It will be possible to manage the cluster as a single system, and to easily schedule user jobs. And, of course, there will be service and support plans available.

Cray has, of course, done exactly what it needed to do to survive. It has been clear for some time that off-the-shelf clusters were going to take over much of the supercomputing and high-availability markets. By offering a Linux-based cluster system of its own, Cray has simultaneously helped to bring that takeover about (by putting a very high-profile name on a Linux cluster product) and to ensure its own future as big iron systems get harder to sell.

(See also: API Networks' announcement of its alliance with Cray to build the SuperCluster systems, and Cray's presentation (PDF format) on the SuperCluster product).

Cray will not have the cluster market to itself, of course. LinuxWorld has brought out a whole set of cluster-related announcements; here's a few of them:

  • Scyld Computing (the company formed by Donald Becker and other Beowulf originals) has announced the first commercial release of its "Beowulf II" code. It includes a number of cluster management tools (added since the preview release) and a wider range of hardware support. Cray is said to be looking at this release for its SuperCluster system.

  • Hewlett-Packard will be using clustering code from MSC for its own cluster offerings, which are aimed at the high-availability end of the market.

  • Compaq has announced a cluster offering at LinuxWorld, produced in cooperation with SteelEye Technology.

  • IBM has been deploying Linux clusters for some time; it's big LinuxWorld Announcement included additional cluster services.

  • Computer Associates announced the beta availability of its "Unicenter TNG Cluster Management Option."

  • SGI is evidently the first to announce the deployment of an Itanium cluster. This one is going in at the Ohio Supercomputer Center; it will be build with 73 dual-processor Itanium systems. They give no date on just when they expect to have that many Itanium processors available...

  • The National Center for Supercomputing Applications and Oak Ridge National Laboratory have put out a press release describing the "Open Source Cluster Applications Resources" (OSCAR) project. OSCAR is a collection of clustering software, all configured to work together and be easy to install; they are calling it a "Supercomputer on a CD." A developer's release is available now, with the full release to follow shortly. IBM will be demonstrating the software in its booth at LinuxWorld.

A first report from LinuxWorld. LWN's staff at the LinuxWorld Conference and Expo has put up a report from the first day covering some of the events that took place there.

Worth noting separately: the "Linus Torvalds award" at LinuxWorld has been awarded to the Samba team. Liz Coolbaugh reports: "Both Andrew [Tridgell] and Jeremy [Allison] kept the audience laughing. They commented that they hope someday Samba will no longer be necessary (which brought cheers) and mentioned that it is now shipping with most Microsoft-based appliances, having become a standard. Jeremy thanked everyone for the award, briefly mentioned the upcoming release of Samba 2.2 and promised, 'You haven't seen anything yet.'"

Hard business models. It has often been said that nobody has figured out how to make money from Linux. That is not true, but it is evident that quite a bit of work still needs to be done in that area. One thing that is worth doing is examining business models that appear not to work; an increasing number of those are coming to light.

Consider, for example, VistaSource. Less than a year ago, VistaSource was split out from Applix as a separate activity; at the time, the hopes appeared to be that VistaSource could go public in its own right. Since then, the reception for Linux companies wanting to go public has been something other than friendly, meaning that VistaSource needs to show that its model makes sense as a money-making business.

Applix has released its fourth quarter results which paint a bleak picture:

For fiscal year 2000 revenues were $50.5 million, compared to fiscal year 1999 revenues of $55.8 million, a decrease of nine percent. The decrease was attributed to an $8.7 million, or 46 percent reduction in revenues from VistaSource, Inc., the Company's Linux subsidiary.

Applix is now actively trying to sell off VistaSource, even though it anticipates losing up to $5 million in the process.

What are the lessons to be learned from VistaSource's problems? Here's a beginning.

Selling Linux office suites is hard. None of the Linux office products has made any real headway against Microsoft's products. In the purely Linux world there is severe competition between ApplixWare, StarOffice (a.k.a. OpenOffice) and WordPerfect. And the free alternatives are getting better every day. A few years ago if a Linux user needed a reasonable spreadsheet, they would almost certainly pay for ApplixWare; now there are multiple free alternatives that work very well.

Auctioning developer services is hard. CoSource.com is VistaSource's effort in that arena; it was launched in May of 1999. According to the statistics on CoSource.com's completed projects page, in almost two years all of 20 projects have completed, with a total transfer of $16,304. It is hard to make a business run on that kind of cash flow.

Collabnet's SourceXchange competes with CoSource.com; it has only completed ten projects, though it has 20 in progress currently. SourceXchange will be doing better due to the fact that its projects generally carry a much higher price tag. Nonetheless, Collabnet has been moving over toward other corporate services for some time now.

In the end, any software project of any significant size will require a personal relationship between those doing the work and those paying for it. It would appear that not too many companies are willing to go out and contract with random developers on the net to get important work done.

Another example: Stormix Technologies, creators of Storm Linux, has filed for bankruptcy protection. One year ago, Stormix had a large, fancy booth at LinuxWorld and gave all the appearances of making a success of its Debian-derived distribution. Now the company may not survive at all. See this article on NewsForge for more on Stormix's trouble.

Between Stormix and Corel, one could well be led to the conclusion that making money from a Debian-derived distribution is hard. Why might that be? The distribution market is crowded in general, of course, so any new entrant is going to have a tough time. But a number of distributions derived from Red Hat Linux have found success. Perhaps the commercial world really does expect to be able to use RPM. Or perhaps everybody who wants to use Debian simply uses it directly, without the need to buy from an intermediary. Whatever the reason, Debian appears to be hard to sell despite being a high-quality distribution.

The one possible exception here, of course, is Progeny, which, by all accounts, is doing well. Progeny, however, has not yet gotten past the beta release of its distribution, so claims about its success are, as yet, premature. Progeny is aiming at certain segments of the market with a high degree of focus, however, and might just pull it off.

All of the companies mentioned above may yet succeed, but at the moment the prospects look dark. Those who are trying to make a business of Linux will want to look at examples like these and be sure that they will not encounter the same troubles. Making a business work has never been easy, even in well understood markets. Free software businesses are operating in uncharted territory; we still have a lot to learn on how to make them work.

Inside this week's Linux Weekly News:

  • Security: A new set of BIND vulnerabilities; DirecTV strikes back.
  • Kernel: Should Linux force ECN compatibility? A wealth of new Linux filesystems; How to sleep well.
  • Distributions: New releases using 2.4, Debian leader elections, Telemetry Box Distribution.
  • Development: The Real Time Linux Foundation, Nautilus Preview 3, Dive into Python, The Linux Development Platform.
  • Commerce: BountyQuest Awards, Linux management products, tons of other stuff too.
  • History: The Cathedral and the Bazaar, Linux Certification, DeCSS
  • Letters: Who's afraid of IBM? Or of Linux?
...plus the usual array of reports, updates, and announcements.

This Week's LWN was brought to you by:

February 1, 2001


Next: Security

Eklektix, Inc. Linux powered! Copyright © 2001 Eklektix, Inc., all rights reserved
Linux ® is a registered trademark of Linus Torvalds