eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
The Enterprise Grid Alliance—a consortium of about 30 vendors and enterprise grid users—on Tuesday will release the industrys first Reference Model for enterprise grids.
The model, downloadable from the EGAs site, includes three vendor-neutral components: a common lexicon of grid terms, a model that classifies the management and life cycles of the components required for enterprise grids, and a set of use cases that demonstrate the requirements for enterprise grid computing in specific enterprise scenarios.
These are the first fruits after a year of work by the EGA. The body came together in April 2004, shortly after Oracle Corp. launched its 10g product line, with the purpose of driving enterprise adoption of grid.
The EGA joined a slew of bodies working on grid computing, including the Global Grid Forum, the Globus Consortium, the World Wide Web Consortium and the Distributed Management Task Force.
The Reference Model itself fills a gap in that lineup, said Paul Strong, chairman of the EGAs Technical Steering Committee and a systems architect at Sun Microsystems Inc.
“At the high level, you have the GGF,” Strong said. “They have the OGSA [Open Grid Services Architecture], a broad architectural description of grids that captures every type of workload and every environment. At the bottom of the stack are organizations like the DMTF that describe components [such as disks, storage arrays and network components]. But theres a notable gap between those.”
Thats where the Reference Model comes in, he said, it being a tool for understanding sets of components that comprise the data center, along with their relationships with each other and what Strong referred to as their life cycles—in other words, how components fit in to existing data center architectures.
“If I come up with a new technology, grid being one, and you want to adopt it in a given data center, it had better fit in with existing architecture, or the data center just wont adopt it,” he said.
Strong said the Reference Models lexicon is aimed at clearing the haze around the collection of terms used to talk about grid.
“The grid landscape as it exists today is somewhat confusing for some people,” said Strong, in San Francisco. “There are a lot of groups and a lot of misunderstanding. If you talk to individuals and analysts and talk about a set of common problems, you come up with a set of terms that are not synonymous. Theres a common set of problems they want to solve, and vendors [use terms such as] grid, autonomic computing, [etc.], that sit around that and solve a set of problems. Theyre broadly similar, but not the same thing. Theres an inordinate amount of confusion in the space.”
The use cases are derived from the EGA member organizations, including customers of Sun and Oracle, as well as from EGAs enterprise user companies, the financial services firm UBS Investment Bank and the United Kingdoms e-Science Programme.
Next Page: The EGAs work seems relevant for enterprise.
EGAs Work Seems Relevant
for Enterprise”> Jonathan Eunice, an analyst at Illuminata Inc. in Nashua, N.H., said the Reference Model has eased the concerns he had about the EGAs relevance when it first launched.
“Theyre doing good work, actually,” he said. “I was very concerned when the EGA first came up. You could maybe still say, Theyve been out there a year and theres just this one Reference Model. But they do seem to have settled into a model of contributing.
“Having a Reference Model, especially when you have a very complicated environment, where there are a lot of moving parts and a lot of change and a lot of confusion over what goes where and what works with what, thats an environment where having a reasonably sophisticated model of the world, and having it in writing so everyone can criticize the same thing … thats useful.”
But its not as useful as having working code, Eunice said, and hed still like to see more in terms of use cases.
As an example, he said, take “Section 7.3, use case example: provisioning server in an operating system. How would I apply the model, what are the components in the reference model that apply to provisioning, what are the pieces that would be put into place, when do we discover the target, when do we load the software onto it, when do we activate the software, when do we report back that the software has been loaded, when do we boot that system thats been loaded?”
These are very low-level steps, Eunice said. But the reason that other standards work, such as XML, is that people have walked through such details.
Still, the Reference Model is a good thing, Eunice said. “Any such document is bound to be too high-level,” he said. “I wish it were more detailed. I wish it had more use cases. But the model I had of [the EGA was] as a pure marketing organization. This is useful output. … Having some documentation about the overall model in which [grid organizations] work, the overall vocabulary in which they work, thats goodness.”
Whats next for the EGA, Strong said, is producing deliverables from four working groups, looking at four issues that are considered inhibitors to the adoption of grid. Provisioning servers and provisioning and management of data, which is at a higher level of abstraction than the storage level, are two issues.
Another issue being worked on is security requirements—of particular interest in an enterprise grid environment, where theres a lot of sharing of computing resources, Strong said, with multiple applications or services installed on a given server or platform.
“You want to isolate those,” Strong said. “And sharing through time: if you have a server or operating system, its at one point part of one server or application. As the load changes, you may want to repurpose the server to work on another application. When you reprovision, you want to scrub and make sure no one has access to information previously on the system.”
Finally, the EGA is working on utility accounting, long a dilemma with grid computing. “We need to map value to cost,” Strong said. “We have to have telemetry or instrumentation to work out which components … which applications theyre doing work on behalf of. That same telemetry enables the ability to pay for services or applications on a pay for use or pay for value basis.”
Check out eWEEK.coms for the latest database news, reviews and analysis.