XML databases are different from traditional databases, and they require a new set of features and metrics for evaluating them. In my last column (XML-J, Vol. 3, issue 2) I talked about native XML database management systems (XDBMS), and I'd like to follow up with how they differ from traditional databases and why this is significant.

When databases were first introduced to the industry eons ago, corporations had piles upon piles of data waiting to be poured into some form of digital storage and management. Not so with the introduction of XML - the biggest bottlenecks today are the piles upon piles of applications trying to effectively use all the available digital information.

Addressing this need is something that is square in the middle of XML's sweet spot. XML is flexible enough to manage any information (structured and unstructured) and extensible enough to support new and changing applications, and it is an Internet standard to cross IT boundaries. For all these reasons, XML database management systems do not usually act as the database of record - although they certainly can - but more often are used for middle-tier data management.

What Is Middle-Tier Data Management?
The typical architecture within an enterprise consists of four layers: back-end systems, business logic, presentation and delivery, and the client application (see Figure 1). Most will agree that XML is the language of choice for delivering information to clients, and few are rushing to convert back-end systems to XML, but the two tiers in the middle present confusion.

Much has to happen to information in the middle tier(s). Most infrastructure is required to:

1. Convert diverse information (structured and unstructured) into XML.
2. Extract the relevant information (a custom view) for each application.
3. Deliver data in real time.
4. Connect to systems not under local control (i.e., systems from another division or a partner).
5. Support changes in applications, partners, and back-end systems.
6. Prevent localized changes from rippling throughout the extended enterprise.

How Is Middle-Tier Different from Back-End Data Management
The two are very different. Consider these distinctions:

Development effort: - Because a back-end database is designed to be permanent, significant development effort is invested upfront to anticipate all future needs. Beyond that, the development is mostly maintenance (connecting to new sources, schema evolution, etc.), and in general the less the better. This isn't necessarily the case at the northern end of the architecture diagram. Development is constant, and occurs every time the organization moves to become more competitive. Whether there's a new partner, a new system, or an additional feature, the data management system requires nearly continuous new development.

Is Middle-Tier Data Management Necessary?
That depends on how much the infrastructure is being stretched to deliver on the six requirements listed above. If data is strictly structured, if batch files suffice, or if agility is not an issue, then it may not be necessary. The alternate ways of delivering XML to clients may be more appropriate, such as flat files or on-the-fly adapters. Flat files are flexible, and of course cheap, but everything is left as an exercise for the reader and there are no enterprise-class guaranties included. Adapters can turn non-XML data into an XML document, but they don't address any of the issues above.

If It Doesn't Walk Like a Duck, It Probably Won't Look Like a Duck
Because the usage models of middle-tier and back-end data management systems are quickly diverging, the feature sets of relational and XML databases are diverging as well. XDBMSes depend on an extensible data model, which is considered heresy to the traditional DBA. Also, simple and standard interfaces, and connectivity options, jump to the top of the priority list.

It should be mentioned, however, that middle-tier data management is not the only way to use an XDBMS. XDBMSes are often used as the database of record or as an embedded database, but those are topics for another day.

Published Mar. 25, 2002— Reads 11,149
Copyright © 2002 SYS-CON Media, Inc. — All Rights Reserved.
Syndicated stories and blog feeds, all rights reserved by the author.

Related Links

▪ Figure 1

About Coco Jaenicke
Coco Jaenicke was, until recently, the XML evangelist and director of product marketing for eXcelon, the industry's first application development environment for building and deploying e-business applications. She is a member of XML-J's Editorial Advisory Board.