Distributed 'Stars' coming out; DB2 2.2 reviving dormant DBMS segment

Software Magazine, April, 1989 by Edith Myers

Distributed 'Stars' Coming Out

To paraphrase Lewis Carroll: the time has come, the blue giant said, to talk of distributed database technology.

It is a fact of industry life that when IBM decides that it is time to talk about a technology, talk takes place.

After working with distributed database technology since the late 1970s with the R Star project in its San Jose, Calif., laboratory, IBM brought the technology into the real world last October with the introduction of release 2.2 of DB2.

"IBM was not the first," said Chris Date, executive vice president of Codd and Date International, a relational database consulting firm, "but when IBM endorses a technology, people pay attention to it."

Date made a point of noting that "there probably is no such thing as the ultimate distributed database system." During DB/Expo 89, San Francisco, he defined such a package as "a system involving multiple sites, connected together in a communications network, in which each site is a database system site in its own right and in which a user at any site can access any data in the network as if the data were all stored at the user's own site."

IBM HAS INVESTED HEAVILY

"Distributed database is an area of heavy investment for us," said Jnan Dash, manager of data systems strategy at IBM.

Dash compared distributed data and dispersed data saying, "both involve data in more than one computer system, but dispersed data is generally inherited [by MIS directors]."

Distributed systems move data to where it is required, under the control of IS departments, Dash said. "Distribution without direction can set an MIS director's teeth on edge." He said careful implementation of a distributed system can avoid the dispersed data situations which are developing because "LANs are much more real now."

To be distributed, Date said, a database system has to be relational. "Distributed database systems are the most significant new developments in the database world since the first genuine relational systems," he said.

Relational Technology, Inc., Alameda, Calif., was the first to market a distributed database product--Ingres/Star--which was introduced in May 1986. Neighbor and rival Oracle Corp., Belmont, Calif., followed with the unveiling of SQL*Star in September 1986.

Both companies chose to use the word (and/or symbol) star in naming their products, a decision observers trace to the code name for the long-running IBM distributed database development project.

Similarly, Cognos Inc., Ottawa, has dubbed its distributed DBMS, introduced in 1988, as PowerHouse StarBase.

Cincom Systems Inc., Cincinnati, brought out version 2 of is Supra relational database system, which executes in a client/server distributed environment, last January.

"When our star component is released," said Marco Emrich, Cincom's senior product manager for database products, "the distributed basis will already be there," indicating that Cincom, like its database brethren, visualizes along stellar lines when it thinks distributed.

IBM NOW INTO STARBURST

IBM, on the other hand, seems to prefer to keep its stars in the laboratory.

While R Star ended as a research project with announcement of the distributed version of DB2, it has been replaced by Starburst which, said Russell Donovan, IBM's database market support manager, is a project aimed at extending the distributed database concept across DB2, SQL/DS and OS2.

IBM seems content at this point to stick to talk among its own products in a distributed world. Others want to talk to IBM too.

"Ingres has no desire to talk to Oracle and Oracle has no desire to talk to Ingres but both want to talk to IBM," said consultant Date.

He noted that a gateway could be created that would make an Oracle site look to an Ingres site like Ingres. However, such a product would be for one-way data transfer only.

"It could not say get ready to go both ways. There are significant technical problems to be solved before this can be done," Date said. "However, there are reasons to do it, and vendors will do it."

RELATIONAL REASONS

Date said all distributed database systems are relational for five reasons: ease of fragmentation; ease of recombination; requests have high semantic content; responses have multiple records; and requests are optimizable.

He said non-relational sites cannot be incorporated into a distributed database system in the sense that its data could be made accessible from a relational interface and could have relational operations performed on it.

"However," he noted, "if the question is, can I do something useful in a DDB with a non-relational database, the answer is probably yes."

Other DDB products currently available include Digital Equipment Corp.'s VAX Data Distributor (released February 1987), NonStop SQL from Tandem Computers Inc. (March 1987) and Sybase from Sybase, Inc. (June 1987).

Date cited two more products that have been promised: InfoReach from the Applied Data Research operation of Computer Associates International Inc., Garden City, N.Y., and Reflex 3.0/DDB-4 from Nixdorf Computer, Inc. of Waltham, Mass. He predicted that other distributed database products will be coming out soon from other vendors.