What are the prevailing attitudes about Grid in the UK? GCJ recently caught up with a Grid pro and veteran Globus Toolkit developer with a great vantage point on both the public and private sector UK Grid efforts for his point of view.

GCJ: In the U.S., the perception seems to be that Grid is still primarily for the research and science realm ... not so much ready for mainstream enterprise IT. Is that the perception in the UK as well?

Parsons: Yes, I give a talk at the moment - most recently to professionals from financial services, oil services, and data management - and I explain how the Grid has been over-hyped for business. This is something these people are already sensing themselves. I'm trying to explain where that over-hyping has come from, where the Grid really is now, and how the Grid will be important to businesses in the long run. Yes, the Grid has been successful in a number of big science projects. And it's been tried out by companies, but not deployed in any great number - unless you're just talking about cluster grids.

There are some complex things going on now. All of the vendors have engaged in Grid, because they see that it may be very important to industry over the long haul. One of the ways that the Grid is being sold in the first phase to companies is as a way of lowering IT infrastructure costs ... getting spare cycles on people's desktops, managing large barns of PCs. And that works - companies like JP Morgan have attributed millions of dollars of savings over the last couple of years to Grid.

But the bigger story - and the vision that people like Ian Foster and Steve Tuecke and Globus always have had - is a Grid that transforms business processes and the way you communicate with other virtual organizations. That story is one that involves businesses investing in the Grid, and investing in changing their business processes. To date, that's the bit that most companies believe has been over-hyped, and they have not yet been convinced to put their money into it.

GCJ: How do you get past that challenge when IT budgets are so tight? What's going to be the tipping point?

Parsons: One of the tipping points - a number of big companies in the UK are taking web services applications very seriously. You can see many organizations going beyond simple web applications and looking at .Net applications, or pure web services applications. It's a very little step now to go from web services solutions to Grid solutions. That's one way it will happen.

The other way is that the Grid has to demonstrate business value. I don't believe that it does enough yet to demonstrate that business value - and that's the fundamental sticking point. It's driven far too much by the needs of science at the moment in terms of what it offers. Enterprise Grid is in a chicken-before-the-egg type of situation.

And there are other types of sticking points. For example, I had a discussion at a financial services event - where large financial services companies said that they'd never deploy Grids beyond their organizations' boundaries. From a regulatory point of view they can't - they're legally not allowed to do so.

However, there are companies that are really interested in building wide area grids that go beyond their organizations' boundaries. Audi is a great example (click here for movie). They see an immediate benefit in sharing designs and simulations with their end users and sub-contractors. That's why they want to get into the Grid now. The more success stories we see like that, more companies will interested in Grid.

GCJ: Do you imagine a Grid emerging, like the Internet?

Parsons: No, I think that organizations will primarily build their own Grids. But there is much discussion about the considerations for connecting Grids. In the academic community, there's been a long running and essentially boring discussion about what the underlying Grid middleware should look like. Should it be vanilla web services? Should it be the web services resource framework? If we're going to get to a point where companies can work together, they've got to agree on a standard - and that standard should be WSRF. WSRF may not be perfect, but it will at least allow companies to join interoperating Grids together.

GCJ: Should the Globus Toolkit be the standard middleware?

Parsons: Yes, but other people can develop WSRF implementations as well. I see the Globus Toolkit being the preeminent implementation at the moment of the WSRF framework - but I suspect all of the major vendors to ship their own WSRF implementations. Certainly within Globus, the strategy is to move beyond the underlying infrastructure and go to the higher level services where we can deliver real value to end users - some of whom are from business and some of whom are from academia.

Even in research and science, I think we're still in the early days. I think the big thing that's going to tip it over is when the large LHC collider (the big particle physics experiment happening in CERN) comes into operation during 2007. In the UK, the Grid's being largely called e-Science. It's been sold as a way of enabling scientists to do science they wouldn't otherwise be able to do. But we're right at the beginning of encouraging new users in the scientific community. There are scientists using it on a daily basis now, but it's really only within the last 18 months that they've been working like that, so we're still in the early use phase.

GCJ: If Grid is still a ways from being mainstream enterprise - how do companies follow it, how do they prepare to ultimately take advantage of it?

Parsons: This is where there needs to be a much clearer dialogue between what business is looking for, and what Grid can offer. One of the things that Oracle is saying - and I agree - is that there's no transactional model for the Grid. If the Grid's going to deliver for a whole load of business processes, it's got to have a clear transaction processing model. Without that, all these enterprise Grid projects mean nothing. Most businesses deal in transactions.

GCJ: The Globus Alliance seems most concerned with research / science Grids. It doesn't seem like the Globus Toolkit has ever had the types of product management that normal products would have in enterprise.

Parsons: I agree. But my view is that GT4 is a very solid release. It's at the level where it could be marketed effectively and used by companies to prove what companies want to do with the Grid and experiment with it. It's a good foundation to build some effective Grid implementations that may eventually be ready to perform in a production context in business. But I don't think even with GT4 that it's ready to perform in a production context in business.

GCJ: So who ushers Grid into the enterprise? Is it a vendor who can plug it into a bigger solution. Or is it the Globus community that have the experience to bear?

Parsons: I think Steve Tuecke's company Univa will have a large role to play in that process with the Globus Toolkit.

One of my jobs is to go out and do pilot projects. And in that process I listen to what companies want from the Grid. And we design bits around existing research projects that will meet the needs of business. We do that from a research context.

Organizations like Univa will be taking those sorts of things forward and delivering them as packaged solutions.

close window