Since its inception nearly two years ago, Data Center Markup Language (DCML) has been the subject of much speculation but few milestones. Conceived as a way of transforming – and simplifying – datacenter management, DCML thus far has been trading on its promise rather than actual results. And for most of 2005, the still-developing specification lay dormant as it was folded into the Organization for the Advancement of Structured Information Standards, or OASIS, a standards body that should add some prestige.
Will 2006 be the year when DCML finally begins to move beyond the formative stage?
"We're going to vet the certification through voting and (through) standards within OASIS by the first or second quarter. At the same time, we'll be trying to push for its use in (test) products that our member companies come up with," said J. Darrell Thomas, an executive with Electronic Data Systems Corp. (EDS) who chairs the DCML technical committee of OASIS.
Specifically, Thomas said reference-implementation software for DCML is expected to be available in 2006. Such products could provide users, other vendors and the standards community at large with the first working examples of DCML's specifications. He cited Opsware Inc. as the most likely vendor to debut such a product.
Goals and Potential
EDS, Opsware and Computer Associates are among a consortium of vendors spearheading DCML. The group of about 20 companies, originally known simply as the DCML
As envisioned, DCML would use an XML-based architecture to describe how datacenter elements relate to each other. The aim is to provide a common infrastructural framework for exchanging data between different components, applications and systems, especially regarding information technology (IT) operations and governance policies. Ultimately, DCML is intended to make it easier for companies to automate such manually intensive tasks as provisioning servers or applying software patches.
Thomas said the consortium is ramping up efforts in 2006 to reconcile DCML with existing process frameworks, including the Information Technology Infrastructure Library (ITIL), a set of practices commonly used for managing IT services. It's unclear, however, what that reconciliation might entail or how it would be accomplished.
Even so, datacenter managers contend that an open standard like DCML is long overdue to streamline the flow of information between servers, workstations, peripherals, software and operating systems.
"If we want to create a business process view of our infrastructure today, we have to get our Web guy, our database administrator, our system administrator and our network manager together to create an engineering document – and as soon as they finish, it's probably going to be out of date. We're looking at DCML as a way to automate those functions," said Steve Hernandez, director of enterprise management and processing services at First Data Corp.
Such a standard would be particularly useful as well in aiding companies with configuration-management tasks and documenting IT policies, especially for disaster recovery, said Hernandez.
The DCML group is basing its process framework on semantic Web standards, specifically Resource Description Framework (RDF) and Web Ontology Language (OWL). Thomas says that structure would better enable enterprises to adapt nimbly to the "evolve-ability" of their datacenter environments.
"If we are able to develop a very useful and keep-it-simple approach using RDF and OWL, I believe DCML might be easily adopted (as a standard). We're vetting that (development) in the group now," said Thomas.
Not everyone agrees with the group's direction, however. Charles Betz, a technical consultant in Minneapolis, Minn., applauds its good intentions but says the semantic Web standards aren't "well baked enough" to build momentum for DCML.
"I think that DCML is a very ambitious project, (but) as a standards consumer I was hoping they would align more closely with existing standards bodies in this space – in particular the Object Management Group, which I think has the capability to execute on something of this nature," said Betz.
RDF and OWL could play a role in advanced interoperability, says Betz, "but the primary need out of the gate is a consensus schema that vendors can choose to support. A visual formalism, serialization mechanism, and mapping-to-modeling tools are all essential to bridge the development/operations divide, and the OMG can provide all of that."
In fact, the DCML consortium may be more notable for the vendors who haven't joined. Microsoft Corp., IBM, Oracle Corp., SAP, Sun Microsystems Inc. and Hewlett-Packard Co. all are conspicuous by their absence. Unless these market-shaping vendors throw their collective weight behind DCML, "you can be pretty sure it won't become an industry standard," says Donna Scott, a research director with Stamford, Conn.-based Gartner Inc.
Even DCML's most ardent supporters are keenly aware they need to drum up support for DCML. First Data is pressuring vendors like IBM to help advance DCML. "Since we're buying a lot of product from IBM, we're telling them, 'Hey guys you need to do this for us,'" said Hernandez.
Thomas said the OASIS committee intends to be more aggressive in its recruitment of large vendors, focusing attention on the rationale behind DCML.
"What we're trying to do is reach out to vendors and say, 'This is not just good for our members, but also good for you as well.' Because we don't want to develop something without having their influence," said Thomas.
Experts say it's probably more likely that large vendors will launch competing descriptive languages for datacenters, which they could roll out in their own products. Sun, Hewlett-Packard and IBM reportedly are already working on developing their own initiatives for automating datacenter management. Eventually, vendors may be compelled to adopt DCML, or some hybrid of it, to satisfy customers tired of doing a "rip and replace" of their architecture every few years.
"Interoperability standards are going to be really critical for enterprises that implement configuration management databases and (want) a more actively managed datacenter," says Scott. "But we're talking years and years and years and years and years."
This was first published in December 2005