Early adopters of a technology get lots of press that drives unrealistic expectations for the technology. How do you avoid that phase?
That's the thing with enterprise grid computing. There's enormous promise in taking a data center from a collection of discreet servers to pools of resources on a network using technologies like virtualization and automation. But it poses an enormous set of challenges.
Grid offers the opportunity for resilience, for efficiency in repurposing and sharing resources and for business agility. When people [vendors] see the promise, they make the promise. Reality is, we can solve one or two of the top business issues facing IT departments in the near term. But [vendors] think they have to sell the Promised Land to get people interested.
Vendors, analysts all have their own vision of 'grid.' Ask 50 people to define it, you get 50 answers. So we need to help customers understand what grid means [in context of their problems], set expectations properly and educate people by providing a simple way of describing grid in the data center. It's a simple set of nouns, verbs and relationships.
All of the promise is there, and people feel compelled to sell the bigger context.
Organizations will not be willing to risk stability for higher compute efficiency. How can grid proponents demonstrate stability?
At EGA we produce prototypical case uses. We want to help provide proof points, show people how to solve real-world problems with grid technologies. If I'm in an enterprise data center, I'm going to lose money if my infrastructure goes down. Data center managers don't want something sexy. They want something that hums in a corner and does its job.
We're having a set of user forums in October. We're reaching out to users to get them to validate that we're reflecting the problems they're facing in their data centers. They're the stakeholders here. No one wants to get stranded at the wrong end of a standard. Middleware, protocols and technologies will come and go before there is consensus. How do you get to that point?
Our goal means we identify where standards are required.
The EGA was started 18 months ago as a consortium of vendors and a few end users to drive standards and overcome these hurdles. We work with a number of the standards bodies, including the Global Grid Forum and the [Distributed Management Task Force].
To make a heterogeneous environment work, you need standards for sharing requirements, shaped by context from the community at large. There is a long-standing IT mindset of one application, running on one server, owned by one company or division. Grid computing takes that control out of people's hands. How do you alleviate fears?
So what's different about grid? It's about sharing. What if I'm sharing three operating systems in one virtualized server? I don't want the admin from one operating system to be able to control another. It's very important to promote isolation. If you're re-provisioning a server, that machine's disk needs to be scrubbed so that the new user can't look at the previous information. It's going to take time.
The cultural barriers exist for a reason. They've created these silos: network people, storage people, database administrators, hardware managers, people to manage the applications. This has created enormous inefficiency, but the silos exist because each of those disciplines has gotten so complex.
The biggest issue with grid is getting rid of the silos. IDC said vendors have a hard time heavily investing in feature rich management tools for technology with an initially small user base. How will that be addressed?
There are a bunch of mechanisms that already exist. If anyone is going to realize the value of grid, you have to hide the complexity, and you need tools to do that. There are higher level management tools that would allow you to manage services [rather than servers]. But you have the same problems with the silos.
[IDC] is right. It needs more time.
But some vendors have tools now that will move application A to box B. And once we trust those mechanisms, we can wrap them around policy. Provisioning now is component-centric. Enterprise grid, in the long run, will allow you to manage the entire data center automatically through policy.
It may take 20 years for people to get to that point. They don't want to automate data centers today. They just want to solve problems. The steps will be small but we'll get there. We've shown people the flag, and now we have to walk them there.
Let us know what you think about the story; e-mail: Matt Stansberry, News Editor