Already a Bloomberg.com user?
Sign in with the same account.
? Tech Support Done Right |
| Yahoo's HotJobs.com is... up, not down? ?
October 03, 2005
Girding for grids
With Red Hat, the leading Linux distributor, hitting balls out of the park left and right, it's no wonder that its business model holds attraction for other startups. The latest I have come across is Univa Corp., a months-old Chicago-based startup that's trying to do for grid computing what Red Hat has done for PC servers. Univa takes the open-source Globus Toolkit, which is used for stitching together vast armies of servers, and packages it as a product suite and service suitable for the needs of large corporations. Univa picked up a huge credibility boost Oct. 3 when IBM adopted the its technology as a core piece of its grid computing offering.
This isn't just an important step for Univa. It's crucial for the future of open-standards-based grid computing--which, in my view--is the best kind. For most of the past decade, grid computing as been the province of academics and scientists. They loop together dozens or hundreds or even thousands of computers and solve compute-intensive problems relatively cheaply. But in the past couple of years, pharma companies, oil exploration firms, and other commercial ventures with big computing jobs have started using grids. Globus, developed by scientists at the Argonne National Lab, is the primary tool used by academics for their grids. It's used in more than 1,000 projects. But, for the technology to be adopted by corporations, it had to be turned into a commercial product with a company standing behind it. Now it has two: One little one and one big one.
In Univa's case, the founders aren't just ambitious fans of the core technology. They built it. Steve Tuecke, Ian Foster, and Carl Kesselman were the original architects of Globus. "Companies need to have a purveyor and a neutral arbiter. We're the gurus, so it makes sense for us to do this," says Tuecke.
Univa plans on delivering its first product before the end of the year. It has already done some consulting jobs to help it learn what corporations want and to test out its technology. It also got $8 million in venture capital in August.
There's a pretty good chance that Globus will become the commercial standard for grids, but it's not a sure thing. A bunch of heavy hitters, including IBM, Hewlett-Packard, Sun, Intel, Cisco, and Nortel, have formed the Globus Consortium to advance the standard. And mighty SAP, the leader in corporate software applications, has demonstrated grid technology based on Globus. But Oracle Corp., the leading database software vendor, is using its own proprietary technology to loop together database grids. It's not clear to me where Microsoft stands. Globus works with Linux, Windows, and other operating systems.
I wasn't able to get Oracle on the phone on short notice to talk about this. They must have perfectly legit reasons for going their route. But, since Oracle is the most aggressive marketer of grid technology, it's too bad that they aren't using Globus as their core technology. Ultimately, it seems to me, that would be the best route to quick and effective grid adoption.
TrackBack URL for this entry:
Listed below are links to weblogs that reference Girding for grids:
? Girding for grids from Telematique, water and fire.
Steve Hamm at BusinessWeek used the IBM-Univa announcement to put in multiple plugs for our business model, grid computing, the Globus Toolkit, and open source software. Link: Girding for grids. With Red Hat, the leading Linux distributor, hitting ball... [Read More]
Tracked on October 9, 2005 07:24 PM
Have you ever tried to use Globus - not for the faint of heart. It is a very complex system with lots of components and many times it just doesn't work.
Now the latest Globus is much better but it has a long way to go to be 'useful' by most organizations. I can see why IBM would support this since it means more $ for their Global Services.
Posted by: Bill Bryce at October 3, 2005 02:00 PM
Globus is only a framework which frankly in most cases is not needed. It doesnt have the core Grid Resource Manager component- One will still have to rely on SGE/LSF/Condor/PBSPro etc .
As a person who has commercially deployed Grids, I would say Globus is an "evangelical software" develped by scientists who were too busy with CERN/TeraGrid style projects and forgot the paying customers - enterprises.
IBM now gets paid to re-engineer the monstor toolkit...smart ...very smart.
Globus is useless for oracle's needs. And as long as Oracle's 10g solves real-world business problems I dont see why it needs Globus.
Posted by: Ritesh at October 4, 2005 03:55 AM
Google? Who cares about Google?
When I go on their site, Iam not seeing a ton of cool/useful stuff for the everyday guy or gal. All I see is another search program that quite frankly doesnt seem to be all that different than Yahoo. If they are going to keep acting like this BIG innovators, why dont they start acting like one and bring out something that people actually are interested in using.
Posted by: Rod Potter at October 7, 2005 01:32 PM
Globus is one of the most over-engineered pieces of technology that I have seen in a long time. It's problem space is so vast, it solves no one thing really well.
Oracle Grid has nothing to do with a Grid as Globus defines it. It is a DB running on a cluster of machines.
People use the word Grid and think it applies to everything. IBM, Sun, and HP are not using globus for anything practical.
The whole thing needs to be redesigned to solve a real problem.
Posted by: John Daniels at October 7, 2005 02:03 PM