Grid computing moves into the mainstream

It's not just research institutions crunching vast amounts of data that are getting into grids

Mention the word “grid” and most IT executives probably think of thousands of CPUs crunching numbers for data-intensive applications such as seismic modelling and drug discovery. Increasingly, however, grids are being considered as platforms for creating more flexible, efficient datacentre architectures.

Companies are looking at using grid technology, which includes monitoring, security and policy-based management, to balance workloads across commodity hardware, industry experts say.

“There is this move away from the siloed infrastructures that have been built up over the last few years,” says Steve Tuecke, who co-founded the Globus Alliance — a research project developing a software infrastructure for distributed computing on a worldwide scale — with Ian Foster and Carl Kesselman, and was responsible for managing the architecture, design and development of the open source Globus Toolkit.

That flexibility, and a resulting increase in use as workloads are spread across a pool of computing resources, is driving adoption of grid technology in datacentres, analysts say. Nearly 10% of all servers sold last year ended up in a grid deployment, compared with less than 5% in 2004 and “just a handful” in 2003, says Addison Snell, research director for high performance computing at analyst firm IDC.

In addition, while the bulk of deployments used to be grids where the focus is on speeding up number-crunching and data processing, nearly half of grid deployments today are more flexible, utilisation-focused deployments.

Financial services firm Lehman Brothers, for example, has used grid technology in some form since the early 1990s. In recent years it has looked at grid as the underpinning of a broader technology strategy that lays the groundwork for utility computing.

“The obvious benefit [of grid computing] is the ability to distribute huge amounts of computations over thousands of CPUs,” says Thanos Mitsolides, senior vice president of Fixed Income Worldwide at Lehman Brothers. “But once you have that ability to distribute computational tasks as services, you realise that you can distribute any services, not just computational services.”

As a result, corporate grid computing and service-oriented architecture (SOA) seem to be maturing together, says Mark Linesch, president of the Open Grid Forum, a new standards body that resulted from the Enterprise Grid Alliance and the Global Grid Forum merging.

“Service-oriented applications and service-oriented architectures continue to be increasingly adopted within the enterprise, and [these] types of applications are perfect for grids because they’re modular and designed to work in scaled-out, distributed environments,” Linesch says. “So you’re seeing grid evolving with service-oriented applications.”

As with many organisations, moving to a company-wide, grid-based platform is still a way off for Lehman Brothers. The first step is bringing a handful of isolated grids, each of which runs on internally developed technology, into an architecture framed by Platform Computing’s Symphony enterprise grid software. Today, Lehman has three grids — about 1,500 four-processor Intel- and Advanced Micro Devices-based blade servers — running the Platform software and supporting interstate derivatives, mortgage and corporate-credit analysis applications.

While the focus is on moving the separate grids into a common architecture, the long-term goal is to create a global grid in which all the resources in the company, whether in New York, London or Tokyo, are centrally managed and shared, Mitsolides says.

“One huge benefit is sharing inside the company, not just sharing of hardware but sharing of services,” Mitsolides says. “Another benefit is that we’ll be able to use less hardware because today our average utilisation is very, very low.”

Grid software from vendors such as DataSynapse, IBM and Platform provides the technology needed to support this goal, Mitsolides says.

“Priorities, scheduling, service-level agreements, synchronous communications, monitoring — all that is common to any good grid architecture,” he says.

Join the newsletter!

Error: Please check your email address.

Tags technologygridmainstream

More about Advanced Micro Devices Far EastAdvanced Micro Devices Far EastDataSynapseGlobal Grid ForumIBM AustraliaIDC AustraliaIntelPlatform ComputingToolkit

Show Comments
[]