In-Depth

In Praise of Data Modeling

Data modeling, that once sleepy demesne of data warehouse architects and hot-shot business intelligence (BI) analysts, has become quite popular.

• By Stephen Swoyer
• 08/29/2007

A data modeling tool is a data modeling tool is a data modeling tool. It doesn’t make for a truism—much less a tautology—for the simple reason that not all data modeling tools are alike.

Thanks to a range of factors—creeping compliance and the adoption of next-gen application architectures foremost among them—the relative pluses and minuses of data modeling tools have become increasingly important to organizations. Indeed: there’s a sense in which data modeling—that once sleepy demesne of data warehouse architects and hot-shot business intelligence (BI) analysts—has become suddenly sexy.

"It’s extremely important. Analytics are becoming even more important [in a business context], and a bad [faulty, incomplete, or un-reconciled] data model can throw that off," said Dr. Wayne Thompson, manager of strategy for data mining and analytics with SAS Institute Inc., in an interview earlier this year. In addition to the other hats he wears with SAS, Thompson promotes his company’s data modeling tool, SAS Model Manager, which provides what he describes as "central, secure" storage for multiple data models. In addition, he explains, SAS Model Manager performs model reconciliation, validation, and performance scoring, too. "SAS Model Manager provides a framework for organizing and tracking [data models]. This is extremely important for regulatory compliance—for things like SOX [Sarbanes-Oxley] and Basel II."

SAS and other players cite growing demand for modeling tools—especially for tools that, like SAS Model Manager, can manage and reconcile multiple, often disparate data models. For a variety of reasons, Thompson and other data modeling proponents say, public and private sector organizations are more hip than ever to the promise—and the importance—of data modeling.

"The merits of data modeling have been well received. [Things like] enterprise modeling, enterprise architectures—this idea that we have to align the information architecture with the application architecture and the business architecture—it’s really helping drive interest [in data modeling]," says David Dichmann, senior product manager for Sybase Inc.’s PowerDesigner data modeling tool. "The catalyst for these architectures is regulatory compliance, which is really pushing people into investigating this. There’s also [merger and acquisition] activity. We’ve got a lot of organizations that we speak with who’ve done a really good job successfully building an integration strategy or a federation strategy, and then they merge workgroups, or they acquire another vendor who’s doing the same thing, and they’ve got to do it all over again."

Sybase recently unveiled a new version 12.5 release of its PowerDesigner tool, complete with expanded connectivity into relational and non-relational data sources, support for a variety of different design environments (including both Eclipse and Microsoft’s Visual Studio .NET integrated development environments) and improved metadata management capabilities.

So why should—or would—an organization tap a data modeling tool from Sybase over competing solutions marketed by more prominent relational database, BI, or application vendors? To a degree, Dichmann maintains, Sybase’s platform neutrality is an inestimable bonus: true, Sybase does market its own RDBMS (Adaptive Server)—along with data warehouse, ad hoc query and analysis, and ETL tools—but it’s by no means a market leader in any of these segments. On the data modeling front, Dichmann points out, PowerDesigner is positioned as a leader in market watcher Gartner Inc.’s "Magic Quadrant."

There’s more here, too, he argues, citing Sybase’s expertise in the high-end data warehousing segment, as well as its growing competence in the ETL segment—both of which further burnish its credentials.

"PowerDesigner is unique. We are extremely open, both on the database side and on the overall platform side," he argues. "We support all leading database management platforms, for example. Not just the majority [RDBMSes], like Oracle, IBM, and Microsoft. We also support MySQL, and we also support a collection of business intelligence databases, like Teradata and RedBrick, as well as our own Sybase IQ, of course." Sybase’s neutrality also extends to development tools and IDEs, he continues. "We would work with just about any development and deployment platform you have. In this release, we’re extending the information liquidity model to extend into ETL and EII technologies. So we can document any ETL and EII, regardless of physical implementation."

All of this begs another, even more fundamental question: why use a dedicated data modeling tool—let alone a best-of-breed data modeling offering—in the first place? After all, traditional methods—such as pen-and-paper guestimation, in some cases—have been successful in the past. According to Sybase’s Dichmann, it’s all about insulation of a sort—in this case, insulation from change.

"The traditional bane of the business intelligence environment is that change makes what’s already running obsolete more quickly. [The] adoption [of enterprise architectures] makes it far easier for you to react more effectively to change, rein in the loss of control that has typically been the death of data warehousing projects after they’ve been successfully launched. If you want to start building an application not just for a workgroup or for a specific project, you’re going to want to do that [build it on an enterprise architecture]."

For this reason, he continues, best-of-breed data modeling will grow in importance as enterprises adopt next-gen application architectures to better insulate their IT assets and business processes from change. "We don’t see the data modeling business as a pure business that’s plateau-ed, but as an essential element to the entire enterprise architecture. Modern, world-class data modeling is key. Tying that implementation together across every paradigm is key."

SAS and Sybase aren’t the only data modeling players on the block, of course. Most RDBMS vendors market data modeling tools of their own. IBM Corp., for example, offers a best-of-breed data modeling tool in its Rational Data Architect, which was developed by the former Rational Software. Elsewhere, vendors such as CA Inc. (developer of AllFusion ERwin Data Modeler) and Embarcadero Technologies (developer of ERStudio) also market enterprise data modeling tools. And other players, such as Allen Systems Group Inc. (ASG) market metadata management offerings designed to interoperate with best-of-breed data modeling tools from IBM and others. It’s a burgeoning—and apparently limitless—market opportunity, proponents argue.

"[ASG’s Rochade] integrates metadata from a variety of sources. It reaches from the mainframe on down to the individual desktop to consolidate [metadata] into a centralized repository," said Scott McCurdy, product manager for repository solutions with ASG, in an interview earlier this year. "This is critical for reconciling business and design models [as well as] resolving [data] definition conflicts. We work with our own [modeling] tool, along with third-party tools, too. And we’re seeing just incredible interest [in Rochade], because of compliance, because of the movement to service-oriented architectures—just because of this desire [on the part of organizations] to become more flexible and more adaptable."

There’s a further wrinkle here, too, says McCurdy: the importance of data modeling and metadata management is really "about having everyone [in an organization] on the same page. That old mentality of siloing or autonomy [between business units, workgroups, and so on] just isn’t tolerable anymore."