GRID COMPUTING: MEGA-HIT OR TERA-FLOP?

Grid computing, which many see as the next step in the evolution of the Internet, shares idle computing resources across both internal and external networks to create a virtual supercomputer on which users can collaborate.

The SETI@Home project is one high-profile but relatively small-scale example of grid computing. It uses the downtime of volunteers' personal computers to search for signs of extraterrestrial life in the form of radio signals. “That's just in case you aren't satisfied with the intelligence we have here,” said Michael Haley, a global executive for grid computing at IBM.

However, tomorrow's grid computing will be done on a grander scale. It will require vast amounts of bandwidth (not a problem), sophisticated software and more reliable and accurate management than that used for networks today.

The technology has many names. The least likely to survive once the concept becomes a widespread reality is probably “grid computing” itself. It is also known as adaptive computing, distributed, high-performance computing, metacomputing or — and this is my favorite — planetary computing.

The various leading proponents of the technology also have names with which they go to market, such as Hewlett-Packard's Utility Data Center, IBM's e-business on-demand and Sun Microsystems' Sun ONE. And they are starting to go to market in big ways.

Initially, grid computing will not find fame and fortune in the telecom space. Initiatives launched recently by the aforementioned companies target the enterprise and scientific communities, many of which are already adopting the technology on their internal networks. In fact, this week IBM announced 10 new grid offerings for the aerospace, automotive, financial, government and life sciences markets. An IBM spokeswoman said the company is in discussions around grid computing with about 10 international service providers and expects to be announcing a relationship soon.

It is unclear what role service providers will play in the grid. Service providers can supply either high-margin, value-added services or commodity bandwidth, said Robert Rosenberg, president of Insight Research Corp.

The speeds at which large players such as HP, IBM, Microsoft, Oracle and Sun are moving to position themselves with the early adopters of grid technology could mean that inaction on the part of carriers would leave them with only one choice: the latter. “And bandwidth is not where the money is,” Rosenberg said.

Grid computing would go a long way toward filling empty pipes and eliminating the bandwidth glut, but service providers should be thinking of playing a bigger role in the technology. “If they just want to serve up pure bits, fast and cheap, clearly this technology will be a driver for demand of fast, cheap bits everywhere,” said Janet Stone, research analyst and Insight Research.

However, Stone said that as computing becomes a utility, a lot of intelligence will be needed in the network to provide value-added services that represent an opportunity for providers. If they don't start positioning for that, “They will just be commoditized and it will be the likes of IBM that consumers and businesses have the relationship with,” Stone said.

Other recent developments indicate that grid computing is gaining momentum. In addition to IBM's grid service offerings, the company announced this week master relationship agreements with grid middleware vendors DataSynapse and Platform Computing as well as relationships with Avaki, Entropia and United Devices. Entropia was involved in the SETI@Home project and United Devices announced last month a deal with Gateway Computer that allows Gateway to offer the unused processing power of its approximately 3000 networked PC and laptop retail display units.

Meanwhile, at the GlobusWorld conference in San Diego two weeks ago, the industry got its first look at the Globus ToolKit, an alpha version of grid middleware. Globus is a research and development project focused on enabling the application of grid concepts in scientific and engineering computing.

Oracle also said last week it would “grid enable” its server products and introduced its own grid developer kit. Grid Technology Partners estimated that grid technology would be a $4.1 billion market by 2005. The first wave of grid computing sales would take place in R&D based on those departments' $18 billion budgets for high-end servers and workstations and $2 billion to $3 billion in high-performance supercomputers.

This activity indicates the dance has begun, but it still leaves the carrier looking for a place to cut in.

IBM's Haley said that despite these recent developments, the industry is just in the first years of development. However, by the end of the decade, Haley expects to see reliable, flexible, policy-based computing “by the drink” that is run under proper security levels and has the appropriate metering, monitoring and billing. “This list of attributes will define grid computing for the telcos,” he said.

Telcos may be tired of jumping every time the next disruptive technology comes along, and it is possible that computing-on-demand may “tera-flop” as bad as the bandwidth spot market. But with the amount of effort going into this technology, they just might want to get behind this one — out in front would be even better.