IBM is closely identified with grid computing through its widely publicised research and development of the technology. However, the computational grid concept - applying the resources of networked computers to solve a specific problem - is already working today, most notably in solving scientific and technical computing problems. IT Week spoke to Kieran Lees, the UK chief of one of the market leaders in grid systems, Platform Computing.
IT Week: The grid concept is reasonably well understood but most people think first of IBM...
Kieran Lees: IBM has jumped in and said "We're going to own this market," but what they don't have is products [so] they're partnering with ourselves and others.
At what stage of development in grid computing are IBM and the other server giants?
They're educating IT as to the way architectures should be designed. They're trying to virtualise the architecture and make it heterogeneous. [But] we can take an application executable and make it distributable.
Grid systems are seen as a way to solve very specific problems - is that fair?
Often it's for a specific part of the business. One customer wanted grid for crash-test simulation because they didn't [otherwise have enough] server resources and it was affecting time to market
How will grid be packaged in future?
I think it will be embedded in the operating system [or even become] the operating system itself. I don't think any application will be developed without it. You'll be able to click a button to grid-enable at source and [the resulting application] will run across an infrastructure based on Globus [the Globus Project software tools for grids]. There won't be a big bang. Grid capabilities will be embedded in developer tools and applications, and companies like ourselves will be able to add value.
A lot of what gets talked about regarding grids is to do with harnessing distributed computing - could service providers and outsourcers benefit?
Yes. To distribute something you have to have a lot of detail about the server, such as CPU usage, memory and so on. If you put that into an Olap database you get tremendous instant [controls] and that's what strategic outsourcing firms want because [many service provider] projects get stymied [by the absence of gauges]. The tools to do that are not there at the moment so our core ability to collect all this data is key. We already have a pretty hefty revenue run rate from workload management and distributed computing.
Can you offer examples of true grid deployments being successful?
It played a big role in mapping the human genome and Phil Butcher, the CIO of the Wellcome Trust, said it helped bring it in two years ahead of plan. In terms of social engineering the implications of that are vast. It's also there in Seti [the Search for Extra-terrestrial Intelligence project].
One of the big concerns with grids is that the ability to access resources from desktops outside the organisation has important implications for IT licensing. How is this issue being confronted?
I don't think its been addressed yet or that the technology has become mainstream enough to require it to be addressed. There will be interesting licensing implications and there's no way software vendors will be able to work per-seat or per-CPU in the future. Software companies that don't address this now will be in trouble.
reader comments