Wake Up and Smell the Data

In Vice President Al Gore's book, The Best Kept Secrets in Government (U.S. Government Printing Office, 1996), Defense Secretary William Perry singles out the Defense Department's Defense Medical Logistics System as an example of a successful outsourcing program. A database management system (DBMS) helped the program achieve this distinction.

DoD used to act as a medical supplies distributor to its uniformed services, which provide health care to 11 million people worldwide. The department would buy pharmaceutical and surgical supplies from manufacturers and store the inventory in warehouses. From there, items were distributed to military hospitals and clinics. The time between ordering and receiving supplies could be excessive, and hospitals would buy pharmaceuticals on the open market to get the drugs they needed immediately. Unfortunately, the hospitals would not always get the best value because they lacked easy access to information about pharmaceutical prices.

To speed up delivery, DoD has eliminated its medical supply warehouses and outsourced the job to commercial distributors. The department was able to close the warehouses by creating an information system called Defense Medical Logistics Standard Support (DMLSS) that enables hospitals to make informed purchases. DMLSS provides medical logistics personnel desktop access to a catalog of supplies available from military depots and vendors. The service runs on an Informix relational database management system.

Before DoD introduced DMLSS, the hospitals "had no ability to [comparison shop] at all," says Army Col. John Clarke, DMLSS program manager. "They'd have had to call up 12 or 13 different manufacturers, who'd be selling aspirin in bottles of 100s or 500s or other amounts. The individual would have had to calculate the price per tablet manually." DMLSS provides a price-per-tablet comparison. DMLSS has reduced the amount DoD medical facilities pay for drugs by 15 percent, saving the Defense Department about $100 million a year, says Clarke.

The system also links the DBMS with an electronic commerce server, which enables DoD facilities to order supplies directly from commercial sources.

DBMS for Dummies

With a database management system, organizations can store electronic information and then retrieve it as needed. Over the years, people who work with data have demanded software that locates data faster, responds to complicated queries and handles non-traditional data such as images and audio. DBMS technology has evolved in response. These days, consumers have four types of systems from which to choose: hierarchical, relational, object-oriented and object-relational. Given the inflexibility of hierarchical and object-oriented systems, however, federal managers will probably find themselves shopping for relational or object-relational software.

A hierarchical DBMS takes the ham-fisted file-cabinet approach to database management-simple lists of data (such as names) are stored in data groups, which may in turn be stored in larger data groups (such as address books). Data cannot be shared between data groups, so the user can perform only limited searches. For example, users can get their hierarchical DBMS to list all employees named Mulder, or all employees who live in McLean, Va., but they cannot get it to provide a list of employees named Mulder who live in McLean.

A relational DBMS, in contrast, can discern relationships between data. The system stores data in rows and columns that can be cross-referenced, rather than in impenetrable data groups. In its most basic form, a relational DBMS consists of a data file, plus access mechanisms. Access mechanisms often include a query planner, which shows how the data can be extracted from the database, and an optimizer, which picks the most efficient route to the data and rewrites the query to perform that operation. Efficient searching is particularly important when a small amount of data is requested from a large database.

Vendors disagree on how relational database management systems are best constructed and delivered. Oracle Corp., for example, markets a stripped down product with minimal processing capabilities-options that would allow the DBMS to process more intricate queries or replicate data, for example, are priced separately. The advantage is that customers can minimize costs because they don't purchase any more database than they need. The disadvantage is that the agency is at the mercy of the vendor on cost if it later needs to upgrade its DBMS to perform new or more advanced queries. In contrast, Informix Software Inc. offers few options with its relational systems-full-blown searching features are built in. However, customers run the risk of paying for features they don't need and the initial price seems high. Other major relational DBMS vendors are IBM Corp., Microsoft Corp., Sybase Inc. and Computer Associates International Inc.

Regardless of how a relational DBMS is constructed, it cries uncle when it confronts data that cannot be organized into columns and rows. A user can store and retrieve complex data like photographs, computer-aided design drawings, maps and sounds in a relational DBMS, but cannot index, search or manipulate it.

The object-oriented DBMS offers an imperfect solution to the challenge of manipulating complex data. It stores data in clusters called "objects" which the user locates and manipulates by following pointers in applications written in programming languages such as C++ and SmallTalk. The drawbacks are that object-oriented systems are expensive to develop and maintain, and most cannot support many users.

"If you look at the architecture of database management systems today, you see the extraordinary ways people went to accommodate the limitations of relational databases," says Michael Keeler, president and CEO of EcoLogic Corp., a software company that focuses on complex data management. Desperate to use complex data with their relational DBMSs, developers united object-oriented and relational database engines, but at first they couldn't get the different storage mechanisms to communicate efficiently.

Enter the object-relational DBMS-software with add-on modules that support complex data in the relational system. Vendors are calling their products "universal servers."

Informix was the first to unveil a universal server, launching its product last December. Its software now offers more than 80 "datablades," each managing a different type of data, including images, maps, sounds and watermarks. Oracle has followed with "cartridges," IBM peddles "data extenders" and Sybase sells "snap-ins" for its universal server products. These add-on modules manage data with attention to its unique characteristics. For example, a module that manages global positioning data allows a user to query the database to find spatial data that intersects, overlaps, or lies beyond or within areas on a map.

Universal servers can handle large quantities of complex data. NASA is looking to a universal server to help scientists perform innovative searches on relational databases containing terabytes of spatial data streaming in from satellites as part of the Mission to Planet Earth program.

Industry experts say universal servers could inspire federal agencies to use existing data in new ways.

"For the last 30 years, the U.S. government has been the largest collector of data in the world," says Jess Worthington, chief technologist for Informix Government. "Now agencies are coming under more and more pressure to use that data constructively and become information providers to citizens." Object-relational DBMS technology gives agencies the ability to determine what new services they are well-positioned to provide citizens, says Toby Younis, director of enterprise systems at Sybase.

Stay up-to-date with federal news alerts and analysis — Sign up for GovExec's email newsletters.
Close [ x ] More from GovExec