Every Manager's Guide to Information Technology:

Extract (7): Relational Data Base

A relational data base is organized so that its contents can be cross-referenced. A major area of development in computer science, software, and business applications, the relational data base is a deceptively simple concept that is immensely complex to implement technically and organizationally. It involves entirely new types of software and extremely high overhead for processing and operations. Despite unsolved problems in many areas of high-volume processing, RDBMS are a key element in creating organizational data resources that can be classified, analyzed, and accessed as if they were part of the indexed contents of an enormous library.

The industry leaders in RDBMS are Sybase, Oracle, and Ingres, though every major computer manufacturer offers the product; DB2 was IBM's flagship data base software through the early 1990s, but the erosion of the mainframe market has slowed growth of installations. All RDBMS have very different design principles, data structures, and performance features. This, of course, means they are incompatible; however, the widespread implementation of the Structured Query Language (SQL) standard for querying data bases allows applications to interface to a wide range of RDBMS.

A much bigger problem is that an estimated 90 percent of the data in large companies is not structured relationally but is maintained in older, hierarchical systems, largely because it is the most efficient structure for high-volume transaction processing or because redesigning and converting the files and the data-processing software would be prohibitively expensive. RDBMS inevitably add overhead from interpreting SQL queries, locating data, checking validity, and synchronizing information. Until recently, this limited RDBMS to either small applications or those that did not involve continuous on-line updating of data. This is changing as computer power reduces the overhead created by the processing delays. But no airline, credit card provider, or large bank could possibly use relational data base technology today to handle reservations, card authorization, or ATM transactions.

The relational model is the mainstream for data-base management, subject to the provisos above. Another approach combines the hierarchical model's efficiency of processing large-scale transactions and the relational model's effectiveness in information retrieval. A hierarchical system cannot handle ad hoc queries such as "list any customer in New York City whose average balance in June or July was greater than $25,000" because this requires parallel processing-a set of linked small and very fast computers, each accessing a small part of the data base. This is similar to using 500 people to handle an ad-hoc query, where each has 30 customer profiles, monthly balance reports, or branch addresses to search. They work in parallel, double-checking with each other at fixed intervals.

Indications are that IBM intends to gamble that parallel processing is the only viable way for the company to recover its failing mainframe business. The company predicts that mainframes will be the data "server" with attached parallel processors off-loading heavy functions. IBM launched its Highly Parallel Query System early in 1994.

The term "computer" originated in the early 1940s to describe the jobs of those who did calculations for the scientists working at Los Alamos on the A-bomb. Machine computers replaced human computers. Now computing is much less important for most businesses than the functions of the librarian and the reporter. The relational model is the most effective effort to date to build machine librarians and reporters.

Until recently, most databases were organized "hierarchically," rather analogous to the filing hierarchy of cabinet, drawer, folder, document, and item within the document. A relational data base indexes and cross-references data at the item level.

Order from:

Order from: