Coercing DBMSs to cooperate - DBMS: Distributed Databases - includes related articles titled 'IBM's distributed database direction,' military's Cals provides distributed DBMS model,' and 'importance of optimization'

Software Magazine, Nov, 1989 by Barbara Bochenski, Mike Bucken

"When Oracle first introduced distributed database capabilities a number of years ago," says Smith, "people weren't ready for it. But lately, in the last year or so, we're seeing an upturn in commitment."

According to Rozenberg, "Ten years ago, nothing much happened with distributed databases because the relational model hand't really 'happened' yet. It was the relational model and Codd's work that gave a big push to distributed database processing. The relational model permitted developers to make their first steps toward a distributed database design.

"The relational model provided the capability to finally implement a distributed database," he continued. "The model itself gives you a lot of flexibility to handle location transparency for example.

"Using SQL, programmers do not have to know exactly where the data resides--on which pack and which cylinder--and what the paths are to the data, The optimizer does it for them. It chooses access paths and goes and gets data. So this is a first step," Rozenberg said.

SINGLE-SITE IMAGE

Location transparency is a major objective of a distributed database system. It is just one of several transparencies that create a single-site image--an important quality of a distributed database system.

A single-site image gives the user the illusion that there is only one database site--their own. With location transparency, users do not need to know at which site a particular piece of data is stored.

"The whole idea of tabularity of data also gives you a lot of flexibility and a lot f independence," Rozenberg said. "You don't have to worry about pointers anymore. There are no pointers between the tables. So if tables are located on the same physical machine or located on two different machines, it doesn't really make any difference, as long as you have a deliberate mechanism that allows you to address both tables."

With the relational model, data can be split up (fragmented) and distributed in a variety of ways. When data is stored as tables (relations), different fragmentations among the sites (vertical, horizontal or combinations of both) are possible.

Fragmentation transparency--another of the transparencies contributing to a single-site image--lets the user behaves as though relations are not fragmented at all. Users are presented with a view of the data in which all fragments are combined.

C.J. Date, in his classic, An Introduction to Database Systems, Volume I, says the twin objectives of data fragmentation and fragmentation transparency together constitute two of the reasons why distributed database systems are almost invariably relational.

While relations are easy to fragment and the fragments are easy to recombine, Date points out, consider what would be involved in performing the analogous functions in IMS or IDMS.

TRANSPARENCY MANEUVERS

Replication transparency is another objective of distributed databases. Data replication means that a given data object may have several stored representatives at several different sites. This provides improved performance and availability. If one site is down, a user can still access the required data from another site.