GCJ: What were the big stories this past year in grid?

Eunice: I'm not sure that there are big stories. Grid is a part of a long evolution, so whatever advances you make in any given year tend to pale beside the tidal wave of change since the late '80s.

I have been, however, tracking more mainstream customers self-identifying as grid customers. That trend may have started earlier than 2006, but certainly for about the last 18 months we've seen an uptick in banks and insurance companies, HR companies-many companies that would not historically have identified with the scientific and technical computing workloads that characterized the first generation of grid deployments. So the biggest news is simply growth in enterprise self-identification as having grids, using grids, and getting value from grids.

GCJ: Greg Nawrocki, the president of the Globus Consortium, has said that he believes 2006 is a make-or-break year for grid computing. Do you agree or disagree?

Eunice: We're talking about a very long-term trend, and so I would have to disagree that we're now in a make-or-break situation. I think grids have made fairly steady progress in industry and business for decades. The question is one of rhetoric: will "grid" continue to be regarded as an industry buzzword describing what people want to do. I definitely do see the terminology "grid" being jostled about with SOA, virtualization, adaptive enterprise and many other terms. So I think the term "grid" needs to be secured now or else other terms like - well, SOA for one - will dominate.

GCJ: What are we to make of the SOA-virtualization-Grid 2.0 alphabet soup? How will it play out? How are these technologies different, or related?

Eunice: The terms absolutely overlap. We've been through nomenclature studies on this, and discovered that you can credibly make the case that any of these is the most original, fundamental technology. That is, you can start from the virtualization point of view and convince yourself that grid and SOA are essentially sub-cases of virtualization. Likewise, you can start in the distributed computing camp and argue that SOA and virtualization are sub-cases of grid.

As I look at it, all of the capabilities involved in grid and distributed computing, virtualization, and simplification of infrastructure, SOA, Web 2.0, and Grid 2.0, improve access mechanisms to the network for people and applications. I think of each of them as important and essentially different. None of them are really interchangeable.

As I said before, I do think that the terminology of SOA and virtualization are ascendant, whereas "grid" is a bit of an underdog as far as "buzz" goes.

GCJ: We hear a lot about standards and the need for standards to push grid and for grid to continue to expand in the enterprise and out of academia. What's your take on standards in grid?

Eunice: Standards help to reduce gratuitous differentiation. Standardization in any interactive or communications medium is very important, even critical. But it's a little bit of a trailing process. Standardization occurs after basic functionality has been set down. And the interesting research cases, the trend-setting commercial cases, precede the standard by definition. They occur because when previous standards have been solidified, base level functionality and interactions are no longer a concern.

I tend to be very pragmatic on the standards question; I think they naturally lag in a rapidly evolving market. As you know, the deployment models for all of these technologies are rapidly evolving, so standards are almost, as a matter of course, going to trail actual use cases.

GCJ: What is your take on the customers you're talking to on the top verticals?

Eunice: If you look at the demographics of grid over the last two decades, you find that HPC cases-the forms of distributed computing that have been used in science, technology, engineering, manufacturing and design-will always be somewhat dominant because there's just such a huge use case.

The interesting change in the demographics is in two new sectors. Firstly, within the design industries-pharmaceuticals, petro discovery, and financial services for example-some newcomers haven't adopted the conventional approach to heavy manufacturing or intensive R&D. They use a new array of applications that are more easily distributed to process larger, more diverse data sets. These new design industry players are coming on hard.

The other new use case that I would say is demographically very important is what I call "financial analytics," or "business analytics." It boils down to judging the relative merits of financial decisions, using a wealth of computation. In one sense, both of these - the new design and the new financial analytics - could be categorized as yet another flavor of high performance computing. But they have quite interesting enterprise dynamics, because they don't just speak to traditional heavy industries. They speak to venture firms, hedge funds, and financial services. So the people that they speak to, the level of the business returns that they promise, and the general use case, I think, are all quite different than what we've traditionally described as HPC.

GCJ: And what are they using-compute grid, data grid, app grid or others?

Eunice: I would say that the second generation of grids, the data grids and app grids, are difficult to gauge since the terminology isn't widespread. People coming at the information management challenge from a grid perspective tend to call them data grids, but others call them data virtualization, or data farms. So all the talk of data grid barely reflects one of the contenders out in the market. The same thing is true for application grids. People call them, variably, Internet applications, app farms or SOA instances. They call them a lot of different things.

And so I'd say the terminology and identification of app grids or data grids is much less developed than that of the compute grid. Likewise, the technology for implementing is based on newer technology, much newer standards, and a new approach. And so less formed in the use case, less formed in the technology, not just in the terminology.

GCJ: I'd like to hear your opinion of the Globus Toolkit and how widely you think it is used?

Eunice: The Toolkit tends to be an implementer's concern. The technologists that I talk to who are at the forefront of making some new things work, they are interested in it. It isn't a business toolkit, nor is it a business capability developer's toolkit. And this is one of the challenges that confronts consortia developing component technologies: it isn't always easy to get executives and directors to understand the value provided by a toolkit.

I think that's a problem with Globus, especially because so many of the distributed databases or alternative grid implementations don't rely on the Toolkit, yet they have a strong commercial profile. So when I talk to people building middleware, who are looking to find a standardized foundation for it, the Toolkit is obviously a part of that discussion. When I talk to people about the business of deploying grid, they're essentially insensitive to the issues of the Toolkit and the underlying technology. It's a mixed story.

GCJ: Any predictions, hopes, or expectations for grid in 2007 calendar year?

Eunice: However the terminology wars turn out, the concept of grid and the standard toolkits and specifications will continue to proliferate. I've used this technology and followed its development since the late '80s, and every year there is a large increase in the number of people that are using the aggregation approach and attempting to push challenging into the distributed context. Users have always been asking themselves what's the least costly way to do it, what's the most forward-looking way of doing it, and what's the most effective way of doing it.

So as the grid approach, the toolkits, and the various grids all continue to evolve, the case for using them is clearly gaining force. At the same time, I think grid is the underdog right now, given how popular SOA and virtualization rhetoric has become. But from a fundamentals point of view, that's somewhat immaterial because the case for deploying grid approaches and the underlying technology is only onwards and upwards.

close window