GCJ: Greg Nawrocki, the president of the Globus Consortium, has been quoted as saying that 2006 is a make-or-break year for grid. Do you agree?

Humphreys: I really agree. I think grid is on the cusp. The promise of grid has always been in the enterprise, and while we certainly see enterprise customers using grid, the applications have been very traditional, in high-performance computing and parallel processing-places where it makes perfect sense.

I believe grid harbors more promise as a platform for transactional applications, and even as an architecture for the enterprise. But from what I've seen and heard, there are very few customers that are using it as for transactional applications. Sure, grid fills a role and it's a useful tool, but the expectations were much grander. And so the expectations for grid either need to be reset, or the technology has to be applied in a way that would allow it to become a platform for enterprise computing and transactional applications, specifically.

GCJ: That leads me to my next question, which is, what is the big driver for grid in 2007?

Humphreys: It's time for grid to become an enterprise platform, or at least a viable alternative. Customers are already making investments in alternative technologies. Virtualization is rapidly becoming a standard tool in the enterprise. And when you consider combining the decoupling capabilities of virtualization with policy-based automation and provisioning tools, you can see that virtualization has the potential to become an alternative to grid. So my concern is that grid could be kind of left out in the cold.

GCJ: There seems to be some confusion regarding SOA, virtualization, and grid 2.0. What do you make of those technologies?

Humphreys: In the case of SOA, the value seems to be in the componentization of the applications. Regardless of whether or not IT professionals "component-ize" their applications, I still think there's value in pooled infrastructure that can be provisioned to applications on demand, whether those are more modular applications with SOA or whether those are more monolithic apps like the current generation. So I think with SOA there is a need for SOI (services oriented infrastructure).

I do think we are trending toward automated, service-oriented, pooled infrastructure that is provisioned, metered, and tied to service levels. And I also think we will see more applications being scaled dynamically to meet end-users' needs. The technology to achieve this could be grid, or it could be an alternative.

GCJ: In your view, what role should standards play in making this happen?

Humphreys: Well, this overwhelming focus on standards raises some concerns for me. I understand that it's a community effort, but it seems like the focus on standards is inhibiting the pace of innovation, because while grid vendors are focused on setting standards, other technologies are advancing. And so part of me just wants to see somebody-a vendor, or a couple vendors-really get behind grid and make it look how they want it to look. And I know standards are necessary, but they can also evolve organically, over time.

To cite an anecdote, I happened to be sitting next to an enterprise customer at a recent grid event. And halfway through the event, he leaned over to me and said, you know, I came here wondering how I could apply this technology, and now I'm not sure I want to use it anymore. It seems like the community is arguing about what grid is, and what defines a grid, and what it should be, but they haven't gotten viable tools in the hands of customers.

So while the industry is still arguing over standards and definitions, customers are saying, well, what am I going to do with this stuff? You guys are still trying to figure this out, and I've got problems I need to solve today. It seems customers are getting turned off by all of that internal discussion.

GCJ: But aren't there lots of customers in verticals actually using grid today?

Humphreys: Yes, I do think grid is being employed by specific vertical industries - its more a question of scale. We see it in oil and gas, financial services, aerospace, automotive, and even some manufacturing. Pharma is another industry that has pioneered the use of grid in new and different parts of the environment. So I have seen some unique uses for grid, but they're still more or less application-based, so it hasn't become a ubiquitous platform in enterprise.

GCJ: Are those verticals you mentioned using primarily compute grid, data grid, app grid, or some other type of grid?

Humphreys: Primarily compute grid is what we're seeing now, and a few data grids. There's two drivers. First, speed, or time to solution: I talked to a customer in the financial services space for whom processing data sets on dedicated hardware took over about 16 hours. When they implemented a grid it took all of 16 minutes to reach that same answer by processing the same data set.

Increased granularity of analysis is the other driver. In oil and gas, companies are able to process their 3-D-seismic surveys at more of a detailed level, and then make better business decisions based on deeper insights into the volumetrics of potential reservoirs. So the drivers are based primarily around that same, heavy compute-intensive application.

GCJ: What new usages are you hearing about beyond compute grid?

Humphreys: We do see storage grids or data grids, if you will, being used. In the medical space There are some fairly visible wins for customers that are sharing information in real time. In terms of app dev grids, I haven't seen anything yet. But as an area of expansion, it should be a leading priority for vendors. Demonstrating that grid is a powerful tool by applying it to a less risky part of the customer's environment would help convince them to apply grid more broadly.

GCJ: What is your view of the Globus Toolkit? How much do you see it being used?

Humphreys: I think it's been pretty widely used by those that are implementing grid. Major grid implementations are typically based on either Globus or Sun Grid Engine. Globus has been a great tool for HPC-type applications. Honestly, though, it seems like it's focused on certain application sets.

GCJ: What were the biggest stories for 2006? Any predictions or hopes for grid in 2007?

Humphreys: Well, there were some visible grid wins, but again, primarily in the compute-intensive part of the environment. That's great and its a very useful tool, but for me I'm interested in the other 90% of the environment. So I'm hoping that in 2007 grid will get beyond compute, and start being applicable to transactional applications. It will go from scheduling and provisioning jobs to scheduling and provisioning applications.

close window