Why DBMS camp wins - data base management system

WHY DBMS CAMP WINS

Information management has split into three major branches, all with the same admirable goals: improve productivity, enable newer and more intelligent systems, and provide higher quality information. The idea is to make computers, and the people who use them, more effective by better modeling the real world with better software tools.

Not surprisingly, these three branches correspond to the three critical information resources managed by enterprises: data, knowledge, and objects. Each branch has a corresponding "camp" in both academic and commercial computing.

The database management camp, represented by vendors, users, and other proponents of relational database management systems (RDBMSs), believes that users need faster and more efficient systems for managing large amounts of basic business data.

The knowledge management camp, represented by the advocates of artificial intelligence (AI) systems, is focused more on data interrelationships in the form of rules rather than on production data.

The object management camp is mostly concerned with improving programming languages to manipulate objects more complex than simple characters and numbers.

By introducing concepts such as inheritance, polymorphism, and encapsulation, the object-oriented camp wants to improve the language itself. It sees the actual data (objects) as a giant virtual file where database management is relegated to the operating system's page fault handler.

Of the three camps, the database camp is clearly dominant in the commercial sector. This, however, is not to disparage the ideas found in the knowledge and object management camps. Many fine ideas have sprung from these areas. But the ultimate commercial winner will be able to incorporate the best ideas of the other two, and bring them to market for production systems.

The database camp is in the best position for three reasons. First, DBMS vendors have the largest installed base and the most experience in production environments.

Second, it is technologically easier for DBMS vendors to incorporate ideas such as rules and objects than it is for AI and object-oriented vendors to write production-quality database systems.

Third, AI and object-oriented vendors have demonstrated almost no interest in providing support for business data management.

EXTENDED RELATIONAL SERVERS

Intelligent databases will be extended relational servers capable of performing both aspects of traditional database work: decision support analysis and backbone online transaction processing.

Such migration has begun today as evidenced by the recent "TP1 wars," where the winners have inevitably acquired IMS/Fastpath optimizations.

Also, intelligent databases will incorporate rule systems reminiscent of those found in systems. Trigger systems, to enforce referential integrity, were the first step in this direction, but these have already been made obsolete by more complete forward-chaining rule systems.

Future systems will likely feature more AI notions, such as declarative rules for referential integrity and backward-chaining, in addition to forward-chaining rule systems.

Intelligent databases will provide toolkits for customizing the server to store and manipulate enterprise-specific objects more complex than characters and numbers.

By customizing the server with user-defined operators and functions, organizations will be able to manipulate complex objects in the server via standard SQL. Eventually, more complex notions such as inheritance will be incorporated into these systems.

BEST OF ALL WORLDS

Enterprises will be compelled to move toward intelligent databases because they will provide a single system with the best of AI, object-oriented, and DBMS technology. Programmer productivity will be enhanced because application programmers will be freed from meticulously reprogramming rules, referential integrity, and special object management into each application.

Intelligent query optimizers that understand computer-generated SQL will enable end-user access, and even reject resource-intensive queries before they lock up the system. And finally, client/server processing will be facilitated in two ways--centrally controlled rules and objects will eliminate maintenance on thousands of clients when a rule or object changes, and network traffic will be reduced by server-based analysis of complex objects.

The future then, will mandate intelligent database servers as the new agenda for information processing. We can also expect new technologies such as object-oriented fourth-generation programming languages, and window-manager independent applications to run on these intelligent servers.

Finally, as connectivity technology and standards solidify, we can expect better interoperability and access between non-relational, relational, and intelligent relational databases.

If SQL is said to "commoditize" relational systems, then intelligent databases which handle data, knowledge and objects will surely again bring even richer differentiation and value to the the information processing market.

Kellogg is the database product manager at Ingres Corp., Alameda, Calif.

COPYRIGHT 1990 Wiesner Publications, Inc.
COPYRIGHT 2004 Gale Group