Deconstructing Oracle's Enterprise Grid

In formally announcing that we've entered the era of grid computing, Oracle has once again brewed a new technology potion for transforming the database world. How radically new is Oracle10g? Here's how Dan Farber sees it.
Written by Dan Farber, Inactive

Once again, Oracle has brewed a new technology potion for transforming the database world. Following on its unbreakable Linux and real application clustering (RAC) rollouts from previous years, Oracle formally announced that we have entered the new era of grid computing.

I'm not sure what's radically new about it, but Oracle has never been shy when it comes to boasting about its technical dexterity or devising catchy names for its latest product initiatives to trump competitors, who have already laid claim to on-demand and utility computing. Nonetheless, the concept should be compelling to Oracle customers looking to upgrade their database infrastructure and take advantage of the lower costs, improved scalability, and dramatically improved manageability.

What is grid computing? Let's start with the more classic definitions. According to Ian Foster, a grid computing pioneer and professor of computer science at The University of Chicago, grid computing has three basic characteristics: coordinating and integrating users that live within different control domains; applying standard, open, general-purpose protocols and interfaces; and delivering nontrivial qualities of service.

Following on Foster's definition, Gartner's resident grid computing expert Carl Claunch defines grid computing simply as coordinating computing resources from multiple owners to handle a single large task.

As an example, IBM, United Devices and Accelrys, along with leading researchers and the Department of Defense, are using grid computing to help find a smallpox cure. Similar to the search for extraterrestrial intelligence (SETI) project, millions of computers worldwide can provide idle compute cycles to the task of analyzing chemical interactions between a library of 35 million potential drug molecules and several protein targets.

The American Diabetes Association will use Gateway's Processing On Demand solution--spare CPU cycles from 7,000 high-end PCs in its retail stores--to help analyze clinical programs and treatments. Gateway rents out its distributed PC farm on the basis of processor cycles used and claims that its grid produces more than 11 TFLOPS (trillion floating point operations per second) during peak capacity.

In a more commercial realm, IBM, Sun, HP and others offer grid computing products and services. IBM, for example, is working with Charles Schwab on an analytics acceleration grid, and Sun has thousands of customers using its Sun ONE Grid Engine.

Oracle's definition of grid computing--distinct from the high-performance computing tasks coming out of academic and scientific communities--mirrors what many in the industry are calling on-demand or utility computing.

Chuck Phillips, Oracle's newly minted executive vice president, said his company's concept of grid computing has three components: pooling resources, virtualizing every layer of the stack, and managing and provisioning resources as a single entity. The net result would be increased be increased efficiency and lower costs due to higher server and storage utilization and automated load balancing and provisioning. "The client doesn't care where the data resides or which computer processes a request," Phillips said.

The concept of a grid aptly describes the mapping and pooling of computing resources into virtual mesh of nodes that can be harnessed on demand. However, Phillips admitted that the company's grid computing initiative is also a key milestone in the battle with IBM because IBM "owns" the term "on demand."

Based on Phillips' description, Oracle's notion of grid computing could just as easily been labeled Oracle On Demand, Oracle Utility Computing, or Oracle N1.

IBM, for example, defines its on-demand operating environment as integrated, open, virtualized and automonic. In IBM-speak, grid computing is a more unique aspect of the on-demand arsenal. According to Irving Wladawsky-Berger, IBM's general manager for e-business on demand, grid computing "is distributed computing taken to the next evolutionary level. The goal is to create the illusion of a simple yet large and powerful self-managing virtual computer out of a large collection of connected heterogeneous systems sharing various combinations of resources."

Unlike clusters and distributed computing, which need physical proximity and some semblance of operating homogeneity, grid resources can be heterogeneous and geographically dispersed, Wladawsky-Berger said.

Oracle is broadening the definition of grids to scale from high performance, vastly heterogeneous fabrics to more general purpose computing grids. In fact, Oracle CEO Larry Ellison said the company is calling its offering Oracle Enterprise Grid to differentiate it from the custom academic and scientific grids. "We are not trying to look for life elsewhere in the universe," Ellison said. "We are doing more mundane things like running SAP, Siebel and our own applications."

During his OracleWorld keynote, Ellison called the Oracle Enterprise Grid the first new approach to enterprise computing in 40 years, and that it would provide 10 times more capacity at one-tenth the cost compared to monolithic mainframe server architectures in use today.

Ellison didn't just unearth grid computing after spending 40 years in the desert seeking enlightenment. Concepts like server and storage virtualization, clustering, load balancing, and unified management are not new. What Oracle has done--based on the details of the announcement--is integrate features such as automated load balancing, monitoring and provisioning on top of the real application clustering (RAC) technology. It's a significant improvement addressing shortcomings of Oracle's current offerings, and geared for more resource-intensive applications.

As David Berlind detailed in his analysis, Oracle is delivering a "grid-in-a-box" solution that takes the complexity out of creating a grid-aware infrastructure, but also requires that you deploy more pieces from Oracle. This shouldn't be news to anyone, given that all the major platform vendors are expanding their stacks and preaching that you'll reduce complexity and cost by going with a single vendor's solution.

"Customers say 'I don't want to get trapped [by a single vendor],' but there are some things you shouldn't be messing with," said Sun CEO Scott McNealy. "The bad news is you can futz with it, and the good news is that you can go in for a tune up or oil change every 100,000 miles. That's what people want--gift-wrapped software. We keep recommending people to stop buying pieces--we'll build it for you and service it for you."

Like Sun, as well as HP, Microsoft, IBM, SAP, BEA and others, Oracle wants to sell a more complete, shrink-wrapped software stack and services. Ellison claimed that all Oracle applications as well as those from SAP, PeopleSoft and others will be able to take advantage of the grid computing features of Oracle10g database and application server without modification. If an IT shop uses BEA or SAP application servers, for instance, with Oracle 10g database, the applications will not work as well as with Oracle's10g application server.

Ellison seemed most impressed with the cost advantages of servers based on Intel parts. He cited that an IBM Power PC server costs an estimated $27,000 per GHz versus a comparable Intel server at $860 per GHz. "If you want to go faster, you have to be willing to spend less," Ellison quipped.

The comparison is extreme and likely not the full story, but the notion that racks of low cost Intel servers are the future of enterprise computing is indisputable. Ellison allowed that high performance, multiprocessor systems won't be extinct anytime soon, but the writing is on the wall. He also said that big iron suppliers like Sun, IBM, and HP have plenty of time to adjust their business to the new world order.

Ellison also brought up a new pricing model for the grid-enabled IT world, foregoing the per processor pricing or number of employees models for an annual recurring fee that allows you to use as much of Oracle's products as you want.

Is Oracle10g a milestone event for Oracle and the industry as CEO Larry Ellison said? No more so than last year's RAC announcement, which has attracted hundreds but not thousands of converts so far. Many other companies have solutions for virtualizing servers and storage or policy-based software provisioning that don't require buying the 10g stack and tools.

Ellison said that within the next three to five years about half of Oracle's customers will have moved to Oracle Grid implementations, but said the big challenge is getting people to think about putting all their eggs in one unified grid basket.

The company has even come up with slogans--sayings from Chairman Ellison--such as, "On the grid, every computer works for a living," or "the power of the grid is greater than the resources."

It will take more than sloganeering to motivate users to move to a new platform. If the economic and performance benefits that Ellison attributes to Enterprise Grid are real, including running applications from competitors unmodified, 10g and its descendents will be an authentic milestone on the road to utility computing, on demand or whatever you want to call it.

Use TalkBack to let your fellow ZDNet readers know what you think. Or write to me at dan.farber@cnet.com. If you're looking for my commentaries on other IT topics, check the archives.

Editorial standards