Grid computing: compelling, collaborative, complex

Get ready for grid.

At least, that’s the message the backers of grid computing technology are sending to smaller businesses, and the numbers seem to back them up. A February report from Insight Research Corp. predicts that worldwide grid spending will undergo staggering growth in the coming years, rising from $714.9-million (U.S.) this year to approximately $19.2-billion in 2010.

“I would say that if it’s not here right now, it’s almost here,” says Dominic Lam, IBM Canada Ltd.’s manager of grid computing. “It’s the right time to talk about the subject.”

More from Dan McLean

To read more of Dan’s articles from The Globe and Mail, click here

But Jeff Kaplan, managing director of THINKstrategies, a consulting services firm in Wellesley, Mass., isn’t quite so enthusiastic. He agrees grid computing is a compelling technology that big companies are embracing. But he believes small and medium-sized businesses (SMBs) are more likely to watch and wait for the time being.

“Large corporations are investigating grid as something they want to acquire and adopt,” he said. “[But] there’s no question that SMBs are still sitting on the periphery when it comes to experimenting [with it].”

So what is grid computing?

Think of it as a networking system that ties information technology resources such as PCs, servers and storage devices into a collective cluster, whether they’re all in the same building or scattered around the world. The idea behind grids is that there’s a lot of computing power in companies today, but it sits idle much of the time while the machines wait for something to process. A grid system ties all these computers together into a sort of virtual supercomputer, divvying up big computing jobs into tiny chunks that are distributed across the network and tackled by available machines.

Grids can also draw on a specialized type of hardware called blades. They are essentially circuit boards of CPUs, which can be installed and removed from today’s state-of-the-art servers and storage devices as more computing power is needed. Blades makes it possible to quickly boost the processing capability of a grid without adding lots of new machines.

The key to an efficient grid computing environment is to set things up so that all the computing capability within a company can be treated as a single resource – one that can be allocated when and where it’s needed. That’s useful, particularly when there’s a demand for tremendous computational capability, such as for year-end accounting or inventory activities. Instead of purchasing additional systems to handle these temporary but intense processing peaks and having them gather dust when the peak passes, a grid provides a more economical and efficient option by calling up the computing power already available on an organization’s network when it’s needed.

Grid computing began life in academia and scientific research as a tool for doing complex and intense calculating tasks. It moved into the commercial world a few years ago, and today sees somewhat limited use in larger businesses, typically for demanding applications and department-specific processes, according to Mr. Lam.

It’s also starting to show up in mid-sized business, where a company doesn’t want (or can’t afford) to shell out for a supercomputer for big number-crunching jobs, but may have several servers and 100 or more personal computers that it could turn into a grid.

“What you need to do is put software on top of the servers to make them work like a grid,” Mr. Lam said.

Software is the brain behind the computational grid, managing the applications and distributing the workload to available machines across the network. The grid’s intelligence comes from its grid-enabled applications and a software-based type of IT glue called “middleware,” such as an operating system like Unix or Linux, that can make a number of CPUs within a dispersed group of machines behave like a unified computer. From the user’s point of view it’s no different than running a program on a single big machine.

Mr. Lam adds that many are finding the technology also creates a more collaborative environment for workers, since all the IT resources are centralized by the grid. “Not only does your department have access to your own stuff … you can look at other resources, too. You end up doing design and manufacturing in a more holistic way because the IT environment enables [that capability].”

So why hasn’t grid technology taken the market by storm?

Complexity, for one thing. Both Mr. Lam and Mr. Kaplan agree building a grid-computing environment usually requires outside help with specialized knowledge. Even large businesses with skilled IT departments typically need to hire a third-party service provider for consulting and integration expertise.

“If it’s a relatively small environment [you’re creating], then it shouldn’t be that expensive to build,” Mr. Lam said.

Mr. Kaplan isn’t so sure. The technology is getting more refined, but he explained that the consulting and integration services required to build and operate a computing grid can still be expensive. The complex operational management required to keep a grid running smoothly, including factors such as prioritization (determining who gets grid resources, when and how) and monitoring (ensuring users are all getting the resources and results they expect), may also prove too much for smaller businesses, he said.

But that doesn’t mean he’s ruling out grids for smaller companies. He touts a managed services approach where, rather than building its own grid, a company buys grid computing “cycles” on a pay-as-you-go basis. Companies essentially rent computing power from someone else’s grid when they need it.

Already such services are available in Canada. In February, Sun Microsystems Inc. unveiled its grid-computing “utility” service that charges an hourly fee as low as $1 per central processing unit (CPU) to use processing power housed at one of its new data centres, including one to be built in Toronto. Through the service, customers can leverage thousands of CPUs for comprehensive processing when they need the power. An Ottawa-based company, GridWay Computing Corp., offers a similar utility service.

Whether they buy grid cycles or set up their own computing grids, Mr. Lam says grids are going to change the way companies think about computing infrastructure investments in the coming years.

“I think the train has left the station,” Mr. Lam said. “I just can’t tell you how long it will take to get here.”

— This article appeared in The Globe and Mail on March 24, 2005.

Related Download
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center Sponsor: Lenovo
3 reasons why Hyperconverged is the cost-efficient, simplified infrastructure for the modern data center
Find out how Hyperconverged systems can help you meet the challenges of the modern IT department. Click here to find out more.
Register Now