Cheap servers, bigger bandwidth drive grid

Grid computing has been around for more than a decade, so why’s it such a hot topic now?

A number of recent trends have collided to make grid computing more attractive than in the past.

One major factor is that servers and server operating systems are cheaper than ever.

“What used to take big iron for big dollars to get big compute resources can now be done by people going to Dell or Sun or HP or IBM, and buying 500 to 1,000 servers at a much lower cost, maybe with a Linux-type operating system, also at low cost, to solve a problem,” said Robert Shecterle, vice-president of marketing with Markham-Ont.-based grid software manufacturer Platform Computing.

A number of companies, including hardware manufacturers and systems integrators are leaping into the grid computing market to offer customers utility computing resources on demand. Instead of buying their own servers, customers could subscribe to a service from one of these utility computing firms and consume only the computing cycles they required. Customers to a utility service like this wouldn’t have systems sitting idle and they wouldn’t have to worry about managing their own computing resources.

One company that’s gung-ho about this form of grid computing is Sun Microsystems, which is setting up a series of grid computing centres across the world.

Sun’s entry into the utility computing market was driven largely by customer demand, said Aisling MacRunnels, senior director of utility computing for Sun. Banks and resources firms were becoming frustrated that they were spending a lot of time becoming experts in setting up data centres and running grid architectures, instead of focusing on their core businesses, she said.

Another driver for grid is the Web and Service Oriented Architecture (SOA). SOA is a collection of services communicating with one another. That communication could involve simply passing data, or it could involve the multiple services coordinating some activity.

SOA is where the true potential of grid computing lies, said Bill St. Arnaud, senior director of advanced networks for CANARIE, Canada’s advanced Internet development organization.

Most existing grid networks are currently internal to organizations, St. Arnaud noted.

“But I think this technology has much greater potential to integrate your business supply chain with your suppliers and your customers,” he said. One challenge companies looking to implement SOA architectures will face is getting carriers to understand the requirements, St. Arnaud warned.

The carrier industry still thinks of networks as one consolidated management facility and that won’t work for grids, he said.

CANARIE, which runs its own network across Canada, has developed a technology called user-controlled lightpaths, that allows users to manage the network as a distributed resource.

So users (in CANARIE’s case these are educational and research institutions) could partition lightpaths, or group lightpaths into a single high-bandwidth pipe. Essentially, what user-controlled lightpaths enable is on-demand networking to complement on-demand computing.

“It’s a new mindset for the carrier industry to say the network can be just another resource like a computing cycle that can be coordinated and managed by the customer as part of their larger grid,” St. Arnaud said.

Related Download
Five Reasons to Take Your Virtualization Environment to a New Level Sponsor: VMware
Five Reasons to Take Your Virtualization Environment to a New Level
Download this white paper to learn how vSphere with Operations Management helps you identify problems and proactively address issues before they affect your end users.
Register Now