Fellows: They're familiar with the name, but that's about it. In our grid adoption research, The 451 Group has spoken with about 60 enterprise grid end-user organizations. There are some enterprise end users who have looked at the Globus Toolkit, but we haven't come across any that have implemented it in a production environment just yet.

Everything is up for grabs in terms of getting the Globus Toolkit effort together. One of the things that plays to its strength is that enterprise end users are looking anywhere they can to use open source technology to offset costs and circumvent the constraints of software licensing. Now that the LAMP (i.e., Linux, Apache, MySQL and PHP) stack is better understood and more widely used, the open source angle should appeal with grids. From what I hear from the users, they find Globus complex -- the iteration is beyond them, and they don't know what to do with it. The immediate opportunity with grids is for someone to define a stack. We've heard too many times from enterprise end users that there are too many grid middleware stacks. They would be happy to work with just one moving forward.

I get the feeling from enterprise end users that there's a slight sense of apprehension about the different grid organizations. I think it's difficult for users to know the difference between the GGF, the EGA and the Globus Consortium, and I get the sense that the momentum is there to be lost at the moment. I think the challenge for all of these organizations is to ensure that the wheels don't come off -- meaning that the early adopters need to be given a simpler path to adoption, rather than a confusing set of choices that they then have to evaluate. That's my concern. The only axe I have to grind is for it to make more sense for enterprise end users.

GCJ: Let's say Globus was a client of yours, and you were advising them to connect with enterprise end users. What should they be doing to get into the mindshare out there - should they follow the Linux model?

Fellows: I don't think grid computing is resonating with enterprise end users. There's no common or shared vision or perspective. We haven't reached that tipping point for grids yet. There's some evidence that we're moving forward. Clearly the technology has been well baked and has come from the right place; I don't think there's any question about it. We're at a hump right now where there's going to be a huge growth in the interest and number of firms in the grid space, and then we'll move into a consolidation phase where the major vendors realize grids are for real. Once these vendors start getting more traction, they'll start buying up the smaller vendors and putting the technology in their stacks. I don't see, in the long term, that there's going to be room for too many substantial independent Grid middleware companies -- but I think one of them might be the Red Hat of the space, if someone can pick up the Globus work and run with it.

It would be ideal if Globus could become recognized as part of one of these stacks -- like a G-LAMP stack. I think it will take something like that to move it forward. If one of the Linux distros were to pick it up and move it forward as a unique packaging, for example. That might work. I don't believe that simply increasing Globus Toolkit downloads is going to do it. The Globus Consortium is shepherding the Globus Toolkit into the enterprise, but there needs to be much better documentation from the major vendors. Several companies have stood up in the last two or three years and said they support Globus -- but where? We're saying, 'show me the products and where it's being implemented,' but no one is really doing a good job of pointing to specific examples and end users.

GCJ: Would you agree that there actually are quite a few Globus Toolkit end users, but that they simply aren't very receptive to sharing their experiences or speaking to the media and/or analyst community? The financial services community, for example, has used Globus on a wide scale, yet rarely speak to anyone about it.

Fellows: You're exactly right. Some of the vendors do have these Globus Toolkit implementations -- and yet they rarely see daylight in terms of people talking about them. IBM has their Grid Toolbox based on the Globus Toolkit, for example, but that's not something that they have at the center of their marketing or communications outreach. IBM had their annual systems and technology analyst event in New York recently, and they had a half day devoted to grids and where they are with things. But no mention of the Globus Toolkit.

GCJ: The theme for this issue of GCJ is 'Networking Considerations for Grid.' In your opinion, at what point will the service providers and networking vendors join the discussion on a greater level? When do Grid and networking converge?

Fellows: Talking to the telecom companies, there are a handful that are trying to grab some leadership space by effectively saying they're going to be grid service providers. None of them have really gotten specific efforts off the ground, however. We've been in touch with more than half of the major telcos, and they all say they have stuff going on, and that the convergence of grid and networking is going to be important for them; however, at the moment, they haven't talked publicly about how that all comes together.

It's disruptive to their own business models -- and they haven't figured out whether it cannibalizes their other services, becomes a value-added offering, or how they'll make money from it. It's all good for users, ultimately, because it will allow them to demand the pay-as-you-go, more cost-effective model. But I think the telcos have been weary of going down that route, because it's closer to the bone. And at the end of the day, the telcos know how to manage networks -- but don't have a clue how to work outsourced enterprise IT services. The whole idea of those companies competing with IBM and Sun for outsourced enterprise IT services isn't going to fly -- with the exception of a couple of markets. British Telecom has a pretty big market in the UK. Deutsche Telecom's T-Systems subsidiary is huge, and they're doing some really interesting things with grids over in Germany, for example.

GCJ: Did you follow the Topspin acquisition by Cisco?

Fellows: I think it signifies that we're in a period where the major vendors are watching this space to see what works and what doesn't. Topspin has been one of these companies that has some really good ideas, so it comes as no surprise that it was an attractive acquisition target for a major networking vendor.

GCJ: The systems vendors have dominated the discussion around Grid so far. What happens if Cisco or another major networking vendor announces Globus support or some sort of common grid middleware support for a backbone?

Fellows: I think it would be a great thing, and I don't know why it hasn't happened yet. IBM has been talking for a couple of years now about how they're working with Cisco to embed grid technology into Cisco gear. IBM and Cisco don't generally say things without meaning them -- so I think we can expect to hear more about this.

GCJ: Are there any additional enterprise hurdles or challenges for Grid that you'd like to address?

Fellows: It's time for a bit of a reality check. Whether we like it or not, when it comes to grids, we're going to hear more about homogeneous and heterogeneous, standard and nonstandard, single domain or multidomain, compute or data configurations. Of course it gets more difficult to separate hype and reality, and it begs the question: Is 'grid' becoming the next 'solution?' The important thing for the industry to focus on is supplying technology and services with classic grid attributes that are distributed, virtualized, standard, shared and loosely coupled, while taking into consideration the new delivery and business models upon them.

Software licensing is emerging as a real thorn in the side for grids. Another hurdle is that the people who recommend buying grids tend to be in R&D, and aren't high up enough in the enterprise decision-making hierarchy. It's not just a technology issue or a vendor issue -- there are all sorts of dynamics within organizations that preclude some of this moving a little bit faster. Another issue is how to deal with legacy applications for those users who are trying to figure out how to leverage legacy gear to work with grid computing. Having to make logical and physical changes to the infrastructure is not an insignificant financial adventure.

close window