In-Depth

Grid passes buzzword stage

Twenty years ago we would have laughed if someone said we would one day have computing on demand. Yet that next logical step in our fast food, microwave age is truly upon us: On-demand computing is becoming a reality.

Better known as grid computing, and sometimes called utility or autonomic computing, the technology can best be described as virtualization. Grid computing is the evolution of distributed computing and refers to connecting disparate computers. “It is a coordinated use of multiple computers and storage devices acting as a single computer,” explained Bob Thome, senior manager of distributed databases at Oracle Corp., Redwood Shores, Calif.

Grid computing differs from its client/server computing predecessor in that client/server required a dedicated, centralized server for each app. With grid, the server is virtualized and can be housed anywhere.

The underlying technologies of grid computing are not new; distributed processing itself has been around for many years. What is new is centralization around an open set of standards and protocols called the Open Grid Services Architecture (OGSA). These standards are what make communication across heterogeneous, geographically dispersed environments possible.

Also new: people’s awareness of grid computing. Not long ago, it was used only in the scientific community; now it is becoming more mainstream due, in part, to the nation’s economy, which has forced organizations to find new ways to make use of their current resources.

For example, instead of assigning a certain number of servers to handle a peak workload at Christmas and then having them idle the rest of the year, grid computing utilizes available resources throughout the year using provisioning, or virtualization. “It allows you to share and pool resources so you can move the right resource to the right place at the right time, dynamically,” said Nick van der Zweep, director of virtualization and utility computing in Hewlett-Packard’s (HP’s) Technology Services Group, Cupertino, Calif.

Traditionally, applications have been silo-oriented. Grid computing is moving away from these silos to a more horizontal approach. This allows firms to add resources on demand to handle peak loads, resulting in better utilization of resources and cost savings.

Because grid computing builds on proven technologies and utilizes current resources, it is not a very risky proposition. In fact, one of the only risks is the cultural challenge of people releasing control to share resources. “If the physical capital assets you’re using are put into a pool or resources, you lose direct control,” said Ed Ryan, VP of products at Platform Computing Inc., Markham, Ontario.

Grid computing mitigates risk by enabling organizations to start small. They can move one app to a grid and then, as that proves reliable and cost-effective, they can transition more apps to the grid.

Kieran Taylor, director of product management at Akamai Technologies Inc., Cambridge, Mass., believes grid computing actually removes risk. “Rather than guess-timate how much hardware or software you might need, we now remove the risk and give you that freedom to innovate,” he said.

The move is on
Grid computing has already begun migrating from the scientific world to the commercial one. Many companies have jumped on the grid products bandwagon; one of grid’s biggest proponents is IBM, which has been inking deals with several grid players lately.

IBM itself offers a myriad of services, research components, software and hardware. But its agreements with other players complement IBM’s middleware stack, said Al Bunshaft, vice president of the company’s grid computing sales in Somers, N.Y.

For example, he said, Big Blue has an agreement with Burlington, Mass.-based Avaki Corp. that “allows us to create global file systems linking together files, not databases,” Bunshaft explained. “It gives us this global file capacity.”

A similar agreement with New York City-based DataSynapse provides IBM with job scheduling capabilities. And an agreement with Akamai makes it easier for IBM WebSphere customers to publish to Akamai’s global network of 14,000 servers distributed across the globe in 2,400 locations.

HP also offers numerous grid products and services as part of its Adaptive Enterprise strategy. The firm’s Utility Data Center, for example, delivers grid capabilities to commercial customers through OGSA standards. In addition, HP offers what it calls Grid Software Infrastructure, which builds on the HP OpenView platform.

Oracle also wants to make grids available to its core enterprise customer base, virtualizing all of its computing resources into one or more pools of resources, according to the company’s Thome. It is attempting to make this goal a reality through a product called Oracle 10g, which includes a database, an application server and an enterprise manager as a suite of grid infrastructure software.

DataSynapse takes a different approach to grid computing -- more of an app-tempered view, said Tony Bishop, chief business architect at the firm. DataSynapse tries to understand the function and service requirements of an app and then fundamentally change the service model and available capacity in the network, he said.

“The industry is coming at [grid computing] bottom up,” Bishop said. “We’re coming at it top down. It’s not about processing over distributed resources; it’s about function leveraging the resources I have.”

DataSynapse offers GridServer, which it classifies as an application operating environment for on-demand grid computing. Based on concepts of autonomic computing, the product was designed to make it easy to extend an application’s architecture for distributed computing.

Avaki has a product called Data Grid that, when added to an existing infrastructure, is said to simplify the provisioning, access and integration of distributed data. The product makes it possible for applications and users to access any data, anywhere, with appropriate permissions.

Platform Computing offers the LSF family of products to aid in grid computing. The family includes multicluster software, a license scheduler, analytics and reporting software. Akamai takes a vastly different approach to grid computing by essentially operating its own grid. In fact, the company claims to provide the largest on-demand computing grid in terms of sheer breadth, scale and processing power in its global network of 14,000 servers, according to the firm’s Taylor.

“Our network will intercept requests and in real time select one of these 14,000 servers that’s optimally suited to accept that request,” he explained. “That’s all done automatically behind the scenes. It’s our secret sauce, so to speak.”

Veritas Software Corp., Mountain View, Calif., is also a contender in the grid space. The company offers various levels and functions of software and services to aid in the necessary steps along the way. Veritas CommandCentral Service, for example, delivers backup, recovery and storage as a service. Integrated with Veritas NetBackup and Veritas Backup Exec, the software measures and reports on IT resource consumption, service levels and costs.

Other companies are joining the push to grid as well. Cary, N.C.-based SAS Institute joined the Global Grid Forum in August, and Sun Microsystems has deals with grid software providers. Other vendors are sure to follow as grid gains in popularity and maturity.

Not yet arrived
Before grid can arrive, however, more steps need to be taken. More standardization would certainly be helpful, especially in the ways these technologies are used. In addition, the information technology industry as a whole needs to start thinking in a more visionary manner.

One thing that would help in this area is for app developers to focus more on building apps and less on business structure, said DataSynapse’s Bishop. “They need to think about building business logic in a componentized manner and with the concept of ‘I don’t care where this gets executed,’” he said.

Oracle is trying to help that along by changing its applications to make them more automated and policy-based, added the firm’s Thome.

But the most important step that needs to happen, according to Bob Maness, senior director of product marketing at Veritas, is to get support for the applications that run in the commercial world. “The more you can integrate across a heterogeneous environment, the faster you can get to a real return on investment,” he said.

Platform Computing’s Ryan believes three key things need to be in place for grid to truly become utility, pay-per-use computing -- which is where it is headed. First, the distributed computing infrastructure itself has to be in place. Second, application migration and adoption of the grid are necessary. “As applications move to that,” Ryan said, “then the value proposition of a virtual environment becomes more realizable.” Third, applications that do not really need a performance grid will come onto the grid because of the manageability of it.

The various aspects of and approaches to grid computing raise the question of whether or not we should differentiate between data grids and computing grids. The consensus from big and small grid players alike is a resounding “no.”

While there are differences in the grids and in uses for the grid technology, they are not significant to end users. “Customers don’t have a lot of affinity to computing grids and data grids, but they are interested in accelerating applications and improving research and efficiency,” IBM’s Bunshaft said.

DataSynapse’s Bishop believes that instead of referring to grids as computing or data grids, we should refer to them as service grids. He advises people to think of the grid as more of a service hub. Computing and data are not separate things, he added; the consumption requirement of an application will include both data and compute power taking place through that hub.

That brings up the issue of how to classify grid software. Depending on whom you talk to, it may or may not be considered middleware. Oracle thinks of it more as management software than as middleware. HP, however, believes grid software is middleware that is built on top of Web services.

DataSynapse classifies grid software as a combination of next-generation middleware and an operating environment. “It’s middleware as long as you explicitly call out that it’s virtualization middleware. It creates one virtual operating environment,” Bishop noted. Platform Computing considers grid software middleware in that it sits between an app and operating system services. It is different from traditional middleware, however, because Platform Computing’s grid products can complement and run traditional middleware, the firm’s Ryan said.

IBM agrees with HP that grid software is middleware. IBM’s Bunshaft likens the technology to TCP/IP, which had to be purchased separately in years past but is now simply part of what we buy. “Grid is heading down that same path,” Bunshaft pointed out. “Now it’s middleware we buy separately; over time we’ll be embedding it into all of our middleware and operating system products.”

Focused development
Regardless of its classification, grid computing promises to change the lives of app developers by freeing them to do their jobs without having to worry about maintaining apps and how they run.

Developers will also be able to do a lot more innovation. “They’ll have more freedom and they’ll do what’s best for their businesses, which is writing in business apps and not managing infrastructure and technology; technology will be self-managing,” said HP’s van der Zweep.

Grid computing, he added, will allow developers to do less capacity planning on an app base. “When they need more or less resources, [the grid] will automatically increase or decrease without having to rewrite [code],” he noted.

Oracle’s Thome suggests that developers take advantage of current standards, like Web services. The reason: A lot of the move to Web services and grid computing will be a parallel effort. “You really can’t pay attention to one and not the other; they kind of go together,” noted Platform Computing’s Ryan.

Grid software should include automated provisioning of resources, the ability to pool and use things without physically having to put them together, and support and integration across heterogeneous environments. It should also be self-contained and include a management element.

One of the most important requirements, according to Ryan, is a scheduling ability. “That’s the compelling thing and core technology that is going to make it all happen.”

DataSynapse’s Bishop agrees. “It needs to be able to use resources that are both dedicated and non-dedicated -- resources available somewhere in the network that have capacity,” he said. “But it needs to be able to know that I need to get off of it for someone else to use it.”

Although grid computing has not yet arrived, it is more than a buzzword. It is on its way to becoming the norm in computing. “This is going to and is changing the way you operate your data center from a siloed approach to a horizontal approach,” noted HP’s van der Zweep.

To arrive at that new norm organizations need to move to a standard way of operating. “If customers are not organized with the right processes to use, they’re not ready to move into a shared world,” added van der Zweep. Only through standardization of processes will organizations be able to truly embrace virtualization technology.

Featured

Most   Popular
Upcoming Events

AppTrends

Sign up for our newsletter.

Terms and Privacy Policy consent

I agree to this site's Privacy Policy.