The recently announced Data Center Markup Language (DCML) may seem, at first blush, like another dull standards effort. It is, however, a fairly significant, albeit underexposed, development.
The DCML effort is a joint effort governed by Computer Associates, Opsware and EDS that is backed by more than 20 other companies, including Akami, BEA, and Marimba. The vendors have tasked themselves with creating an XML-based language that facilitates communication among the various devices in a data center.
The ability to communicate in a standardized format is vital for the future of utility computing. "In general, what the driver is to DCML is the need ... to rationalize IT costs and shift to automation and utility computing," says Eric Vishria, Opsware's director of product management. "One of the things people realize -- and you can see it by the number of companies behind the DCML effort -- is that there are some technical limitations today in achieving that vision."
In a nutshell, DCML inventories what hardware and software is present on a system and reports it to other devices. It can also provide information to other devices and their minders on the status of that device. For example, a server may let others know that it is offline for whatever reason on every other Thursday.
"In this new age of utility computing, you need information to do things like rapidly repurpose servers and reallocate resources," says Vishria. "That's where DCML comes in."
Providing all these "descriptions" isn't easy. Not only must servers, network equipment, disk drives, and other elements be taken into account, but the software that drives these devices must be considered as well. And the prize for dealing with this complexity? A far more elegant IT infrastructure.
"You are talking about a common language that is going to describe how all components of a data center are going to interrelated and work together," says Larry Shoup, vice president of product strategy for Computer Associates. "That's a complex problem."
At this point, DCML has a long way to travel before it is a commonly accepted standard. Corporate infighting will likely be one of the first battles it must win. As history has shown, enterprises that own a significant piece of a marketplace are disinclined to make it easier to match their gear with those of companies whose marketplace position is more tenuous. Bigger companies typically do not come on board until standards gain acceptance and they are left with little choice but to comply. Thus, Hewlett-Packard, IBM, Microsoft, and Sun Microsystem are not in the consortium.
The DCML faithful aren't shy about sharing their goals for the standard. Their vision for DCML is that it will do for the data center what HTML did for the browser and what IP did for the overall network, Dave Thomas, chief technologist for EDS Hosting Services and a co-author of DCML told ServerWatch. Thus, DCML has the potential to work far beyond the data center and become part of the overall effort to share computing resources across corporations, in endeavors such as grids and clustering.
Standards generally arise out of need and are typically ambitious and complicated. Based on recent developments in the hardware industry, this alone is one reason to follow DCML as it evolves. During the past year, a host of major companies have made significant announcements about new ways of leveraging computer power. A cornucopia of terms have been used to describe them -- grid, utility, and pervasive, being among the most widely used.
The one thread they all share is the capability to delegate the resources of an individual computer on a far more flexible basis than before. Thus, the machine -- and, by extension, the IT holdings of the entire organization -- can be used far more efficiently.
There is an undercurrent of skepticism about these approaches, however. Some enterprises feel that vendors, eager to shake off the lethargy of the past few years, have over-promised on what they can deliver. Perhaps the marketing departments are ahead of the engineers. The big questions on the minds of most management teams are: Do these terms refer to truly new ways of leveraging the data center, or are they simply tweaked business models? And does software exist that can truly virtualize computing resources horizontally between machines and applications?
And that is the crux of what makes DCML so potentially valuable. It clearly isn't the only effort at creating a language to push true utility computing. But, as a standards-based, non-proprietary effort, it may be the best stab yet at fulfilling the engineering, not the marketing, vision.
The DCML Organization's goal is to release a public specification for comment by the end of the year and submit a proposal to a standards body in early 2004. It is undetermined to which standards body DCML will be submitted. OASIS, an international e-business body, is a good possibility, Vishria says.
Products that follow pre-standard versions of DCML may be available as early as 2004.