New Data Transformation Model Gains Momentum

By Maria Trombly | Securities Industry News | January 12, 2004

Think of a bicycle wheel. You've got the rim, you've got spokes and you have a hub at the center that all the spokes are attached to.

Now, imagine there was no hub and that the spokes went straight across. And that it wasn't enough to just connect a point on one side of the wheel with a point at the opposite side with one spoke. No, each point had to be connected to several others--a kind of a fan of spokes coming out of every point. That wheel would now be a tangled, solid mess of wires.

A tangled mess is what we have now when it comes to transforming data from one standard to another. The hub-and-spoke model is the promise of ontology-based transformations, also known as vocabulary management.

"The next wave of innovation in transformation will be based on vocabulary management," said Roy Schulte, an analyst at Gartner.

Ontology-based transformations are already available in commercial products, he said, but were used in less than 1 percent of integration scenarios last year.

"We've got some big changes coming about how we do transformations," he said. "In two or three years, everybody will be talking about it. In five or seven years, everybody will be doing it and saying: What's the fuss?'"

Ontologies represent a major step towards creating a semantically aware enterprise. In other words, a company's systems will not just blindly pass data back and forth but will know what the data means. Thus, a company will be able to compare and compile information from different systems and from different business processes.

According to Schulte, ontology-based transformations will eventually reduce the development time needed to implement integration, lower costs and reduce custom coding. "We are just coming out of the advanced technology stage and starting to put this to commercial use," he said.

Today, there are two ways to create a usable ontology-based product. One is to create a standard data dictionary, a universal standard that includes every element that the company uses in any of its databases. Messages are translated into this common language and then translated out again.

Another way to use ontologies is to create that central dictionary, then only use it to build transformations. This speeds up the integration process and makes it possible to update transformations quickly, but doesn't eliminate the tangle of connections.

One company taking the latter approach is Contivo.

According to Dave Hollander, Contivo's CTO, and a co-inventor of the extensible markup language (XML), ontologies can play a key role in the design phase of a project.

"We use it in the design work, and compile it into code that can work within your existing infrastructure," he said. "Which is a huge savings in terms of getting changes into production very quickly. This allows you to work in an existing infrastructure that's not semantically aware."

It can also save time during a transformation itself, because messages don't have to be translated to the central dictionary first, he said.

Contivo is currently working with a multinational broker-dealer with more than $300 billion in assets, said Hollander. The customer has decided to use XML as a basic exchange standard for its business lines.

The downside of not using ontology-based transformations is huge, because the neutral central state, the hub of the spoke system, is critically important, said John Parker, VP of financial services at Contivo rival Vitria Technology.

"If I'm a retail broker," Parker said, "I'm taking orders in from the Web and through my network of branch offices and through sales desks in my banking partners offices and I have 1-800-BROKER where I'm taking orders over the phone. If each one has its own messaging protocol, there's no place in your system where you get global visibility of your retail traffic. And that's important from a business management perspective, but also from a legal and compliance perspective."

Parker also pointed out that in a typical financial transaction, a message will undergo somewhere between 10 and 20 transformations. "If you use the direct transformation methodology, then you lose the ability to get an end-to-end transaction life cycle view. If you have an intermediate representation, I can track a trade as a trade--it's not just a message, one of 20 messages associated with a trade. That means I can identify systemic problems and keep problems from happening again--that's what STP is all about."

An ontology-based approach to message traffic also allows the creation of universal rules about how data is handled. Companies can create common glossaries, such as "MA=Mass.=Massachusetts." Business rules can be applied to at the hub in order to funnel messages in the right directions. And certain data types can be labeled as confidential. Then, when any of the above needs to be changed, it gets changed only once, at the hub. Then, as new applications are written, they can accept and send messages directly in that neutral, common format, Parker noted.

For Vitria, that common language is grounded in ISO 15022. The standard, if widely adopted by Wall Street firms, would make ontology-based transformations extremely attractive, said James Hartley, chief technologist at the Financial Information Services Division of the Software & Information Industry Association.

Hartley, who is a co-architect and co-editor of the MDDL (market data definition language) specification, is watching ISO 15022's development closely, and hopes that eventually MDDL will be a subset of 15022.

According to Hartley, it will be three years or so before 15022 is all-encompassing--and thus useful right out of the box as an enterprise ontology for a Street firm. Then it will take another year or so to go through the certification process, he added.

"Then, by having a common ontology or this common set of terms, you can write business rules to that common set of terms and it simplifies everything," he said. "That's very valuable."

Parker calls ISO 15022 the most important emerging ontological standard for financial markets. Today, he said, almost all Street firms use point-to-point transformations because that's the only technology that's been available. However, he estimated that 95 percent of Street firms are now investigating ontology-based transformations.

"When Swift changed to 15022, about two years ago, they made a very strong recommendation about how you handle financial messaging," he said. "And their recommendations included a separate business rules depository for transformations and a separate data dictionary for common business objects. So all the vendors started shifting in this direction and the customers began to shift in this direction as well."

One Vitria customer is RBC Investments. An RBC executive said that ontology itself can represent a competitive advantage.

"Defining corporate enterprise standards is key to competitive differentiation," said Morteza Mahjour, SVP of enterprise information and technology services at RBC Financial Group.

XML: Provides a surface syntax for structured documents but imposes no semantic constraints on the meaning of these documents.
XML Schema: A language for restricting the structure of XML documents.
RDF (Resource Description Framework): A language for creating a data model for objects and relations among them, providing simple semantics for the data model.
RDF Schema: A vocabulary for describing properties and classes of RDF resources, with semantics for generalization hierarchies of such properties and classes.
OWL (Ontology Web Language): Adds more vocabulary for describing properties and classes, as well as relations among classes.
UML (Unified Modeling Language): Allows software engineers to design an ontology by organizing object classes in a hierarchy and creating relationships among them.

Source: IBM