May 7, 2007
After commenting last week on the shutdown of Grid.org, United Devices' altruistic site featuring various volunteer projects, I received an anonymous e-mail alerting me that there was more to the site's shutdown than meets the eye. This mysterious messenger suggested that UD had laid off 90 percent of its staff and that Grid.org, like those employees, was simply a casualty of UD running out of money. Upon further investigation -- a phone call with UD president Ben Rouse -- it turns out my "Deep Throat" was right ... sort of. At the least, he wasn't entirely off-base.
As it turns out, UD did eliminate a significant portion of its sales staff (around half of it), as well some folks in the marketing and technical departments. However, these "strategic changes," as Rouse calls them, were not a matter of running out of money, but rather a move to focus the company's resources around its push toward selling "application-centric virtualization" and datacenter solutions. The sales force, he said, had been comprised of people with HPC backgrounds, and that just isn't the focus around UD anymore. "Honestly, we don't need a lot of marketing or pure sales support to continue to grow our HPC revenues," said Rouse. "If anyone is out there in R&D land and is considering implementing a grid, they find us." UD is currently looking to fill this void with veterans in the field of selling to telcos and managed services providers, as well as with in-house technical experts in domain-specific areas, such as SAP applications.
While this explanation certainly makes sense from a business standpoint, there's always something a little disconcerting about seeing people lose their jobs, whatever the reason. With this in mind, I would just like to wish everyone affected by this decision the best of luck in whatever they do next. Although hardly definitive in nature, my dealings with UD have always been amicable, and I'm sure cutting loose a large number of employees wasn't on the top of anyone's list.
The shutdown of Grid.org, Rouse explained, was just another part of this reorganization -- although I'll bet it was a far easier decision to make. Maintaining the site was far from resource-intensive, he said, requiring a whopping one-quarter of a systems administrator, but the resources required to find good projects, build applications, etc., did begin to add up. Quite frankly, commented Rouse, Grid.org was never intended as a revenue stream, but rather as an opportunity to showcase the power of large-scale distributed computing -- which it undoubtedly did. With this in mind, it was decided to cease operations instead of spending time and money searching for the next project. Rouse said, and I concur, that those interested in volunteering their computers' resources are better off looking to IBM, for example, who, resource-wise, is in a far better position to handle this kind of an effort.
Market-wise, however, the story is of another stalwart grid vendor moving its focus away from HPC and into the datacenter. In 2006, Rouse said, the company escaped its "life sciences pigeonhole" and saw HPC revenue spread evenly (at 20 percent apiece) across the life sciences, telco, manufacturing, and oil and gas verticals, with an additional 20 percent spread across a variety of others. More notably, though, was that HPC revenues comprised less than two-thirds of the company's revenue, with the rest coming from datacenter customers. Ninety percent of that datacenter-driven revenue, Rouse said, was on the managed services side of things, which is why UD is now "maniacally focused" along this front. Driven by managed services sales, Rouse said UD expects more than 50 percent of its 2007 revenue to come from datacenter solutions.
The company's product portfolio mirrors this datacenter focus, with HPC being just one of three solutions sets, the others being "Data Center Virtualization" and "Managed Services." Each one of the solutions available under these headings, said Rouse, is based on the Grid MP framework and features a customizable combination of at least one more of UD's products -- Synergy, Insight, Reliance and Affinity -- the latter two of which are designed specifically for the datacenter.
However, like DataSynapse and Platform, United Devices, Rouse said, doesn't consider this so much as a new focus as much as it does a natural evolution. The inherent scheduling prowess of the grid architecture is well-suited to tackling datacenter issues, especially those involving virtualization. These three vendors in particular, he added, are at a distinct advantage in this market because of their well-established user bases, many of whom would love to bring the reliability of their HPC systems to bear on business-critical applications. The aforementioned companies, said Rouse, have been evolving along these lines for two years, and the path is well-documented in product announcements and partnerships. "I don't think this should come as a surprise to anybody," Rouse stated.
So, there you have it: Not only the real story behind the demise of Grid.org, but, more importantly, a look into what this migration into the datacenter means for traditional grid computing vendors. It's not always pretty, and it doesn't mean HPC users will be left in the dust, but it does mean that priorities and solutions will shift, and the definition of grid computing -- whatever that might be -- will never be the same. I've said before that the technologies being touted by smaller companies like GigaSpaces and Appistry aren't your father's grid computing, and now, it seems, even the "fathers" in this scenario are trying to change that old-style perception, as well.
Moving onto this week's issue, I first and foremost need to point to my interview with Charlie Catlett, former TeraGrid director and now CIO of Argonne National Laboratory. Charlie shared many insights into the TeraGrid project, including, interestingly enough, that his stepping aside was not only the right move for him, but for the project as whole as it moves toward a more distributed, democratic leadership model. Essentially, it's a lot of good insights from a man with plenty of experience across the grid and general HPC communities.
I also would suggest reading the rest of the Special Features section, which includes a firsthand explanation of Microsoft's SecPAL project, a look at the future of enterprise grid and the announcement of Sun's making Network.com available globally. Beyond that, other articles of interest might include: "Paremus, QuIC Deliver Distributed Analytics Solution," "Univa Launches Open Source Cluster Solution," "Callidus Makes On-Demand Solution Available in EMEA," "OASIS Forms WS-Federation Committee" and "SDSC Selects Brocade SAN Solutions."
Comments about GRIDtoday are welcomed and encouraged. Write to me, Derrick Harris, at firstname.lastname@example.org.