About a year and a half ago, I met with several members of Sun's data warehouse marketing group at one of the data warehousing shows. At the time I was in the process of working on our first data warehousing/decision support study, Large-Scale Database Solutions, and I wanted to learn more about Sun's future directions in that market. I also wanted to check the validity of the revenue numbers I was attributing to Sun.

When the Sun people saw my numbers, the immediate reaction was that our estimate was too low. In the course of the discussion, it came out that Sun was concentrating its marketing effort on the very low end--small data marts and the like. Since at the time I was interested only in the high end of the market, they grudgingly admitted that my (relatively low) estimate for Sun was on target. The clear message from that discussion was that Sun believed that the high end of the market was too small and too crowded with competitors and that the way to make money in decision support was to sell jillions of little UNIX-based data marts at fifty grand a clip.

It's kind of fun to watch a big company like Sun eat its own words. While Sun would certainly like to sell boxes in the jillions, it has discovered that in the data warehouse space, big is beautiful!

The fact is that Sun's high-end Starfire system is going great guns as a large-scale data warehouse server that can hold its own against any of the biggies in today's market. Sun appears to have discovered, along with most of the rest of the IT industry, that client/server is often synonymous with chaos and that centralized computing, especially when big databases are involved, is okay after all.

Starfire is the latest incarnation of a system originally designed by Cray Research who gave it the inspired name "CS6400." The 6400 was based on technology which Cray had obtained from the ashes of Floating Point Systems and Celerity Computing, two high-performance computer companies long since passed into cyberheaven along with about fifty other HPC wannabes. Cray took that base technology, added a significant dollop of its own goodies, and managed to build the first SMP computer that could scale beyond eight processors without virtually grinding to a halt. In fact the 6400's performance curve was darn near linear up to its maximum of 64 processors for many applications. Cray tried to position the 6400 as a platform for business computing, but couldn't shake its techie image and didn't meet with a whole lot of success.

Enter Ed McCracken about the time that Cray was going to announce the next generation 6400, to be christened "Starfire." Ed, then the CEO of Silicon Graphics (SGI), lusted after Cray and wound up acquiring the company, thus giving SGI a series of migraines that have yet to be cured. One of those migraines was thought to be the 6400 program. The idea of supporting a system that ran Solaris, for Pete's sake, was considered an anathema by SGI's powers-that-were, so they sold the thing to their detested neighbor, Sun Microsystems, for a few bottles of beer, no doubt thinking that Sun would now get migraines too.

Whoops, big mistake. The guys from Oregon (Floating Point), California (Celerity) and Minnesota (Cray Research) really did know how to design a reliable, state-of-the-art, high-performance system good for business applications! Whether or not Sun knew that beforehand or discovered it after they got it is a moot point. I'd say McNealy is either very smart or very lucky--probably a bit of both.

Starfire is today Sun's flagship data warehousing platform. Word is that they can't make 'em fast enough and that the customers are standing in line. Of course, Sun has to give the impression that Starfire is its very own, so some marketing genius added the tag "Ultra Enterprise 10000" to its name. Nonetheless, a local pundit characterized the system as the "biggest, baddest computer ever to hit the commercial marketplace." I guess there is something to that, since Sun is in the process of ramping up production to the tune of 150 systems a quarter. At an average of $1 million+ a pop, that ain't too bad, and that's just the basic system price. When you start adding in all the peripherals, networking, storage, software, workstations, etc., that accompany a big system sale, you are talking major bucks.

One of the major benefits of Starfire is that the system can be partitioned into different logical domains. This would allow, for example, a user to run OLTP in one domain and a warehouse in another domain. Our studies have shown that lots of users out there would like to have a single platform running both types of applications. Many believe that it is easier to manage a single system than multiple systems and that there are fewer problems with data transformation.

Here are some stats supplied by Sun:

  • 25 percent of Starfires sold are sold for data warehousing.
  • 60 percent of Starfires are running some type of DSS application.
  • 15 sites with databases over 1 terabyte.
  • Average system has more than 32 processors.
  • Oracle and Informix are the most popular RDBMSs with DB2 coming on strong.
  • In 1997, Starfire set at least one benchmark record every month.
  • If Starfire were a mainframe, it would be ranked number two behind IBM in MIPS shipped in 1997.

Now if the damn thing would only run NT . . .

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access