Replicating What Is Different

We have talked in this book about replication between two separate Oracle databases. Now we must tackle the task of replication among databases that might be totally foreign to Oracle in every way. The ideal form, where a large network of connected databases act as one database, replicating one another’s data, is a distributed database.

A distributed database is singular in that although data physically exists at many sites, it is still one database at the highest level that is simply distributed physically, as in Figure 8.10.

Figure 8.10.  A distributed database appears to users as one database.

Oracle solves the mystery of the distributed database in a twofold way. Oracle first offers a Transparent Gateway to the foreign machines, so they all appear to be Oracle databases. This also involves the purchase of SQL*Net to make different network protocols appear the same. After these tasks are accomplished, proceed with replication in the same way as if the databases were all Oracle databases.

If instead you only need consistency, at day’s end, you don’t need to constantly refresh transactions but can defer them for propagation to other nodes of your distributed database. To do this, you need to store logs of your transactions and then at a given time use the log information to refresh remote nodes. This type of replication is Asynchronous.

Obviously, Asynchronous replication is less sophisticated in that each transaction does not need to be propagated to other nodes of the distributed database but instead to a log file. Because of this, a two-phase commit will not need to exist for a transaction to complete. This makes for a faster transaction rate.

With Asynchronous replication, we do have the added worry of our transaction logs. Even though they fill up only with rowids, timestamps, and other logging information, after many transactions, log files can become large. Therefore, Asynchronous replication can take up more space. But today, with speed more valuable than space, this is not a major concern. A megabyte of storage can be purchased for a rupee.

One catch with Oracle’s replication, even if you have a Transparent Gateway, is that you need the IBM Data Propagator to move data from the mainframe database back to Oracle, either synchronously or asynchronously. This is vital in configurations where a mainframe is still the central computer of an operation with new Oracle departmental servers existing to perform local processing.

Features of Oracle Replication Services

After you purchase a Transparent Gateway and the Replication Services software from Oracle, you gain many advantages in replicating Oracle data with data from a remote database. In the most common case, these remote databases will be mainframe databases. Here are some of the features of Oracle’s Replication Services:

•  On the Oracle end of things, if you use Windows NT or UNIX, your replication efforts are coordinated by a graphical front-end. This is far easier to use than a line-mode interface that forces you to execute commands. This includes automatic error-recovery.

•  Column and Row subsetting is available. This is very important when interfacing a mainframe environment to an open system. Many times, mainframe columns are not as normalized and carry old-style keys. For instance, you might have a column on the mainframe called flight_code, which is 17 characters representing three different codes:

Position 1 - 2         Airline Code

Position 3 - 10        Coupon Number

Position 11 - 17     Reservation Code

Using Oracle’s Replication Services for a distributed environment lets you map this one column on a remote mainframe to three columns in the Oracle database.

•  Oracle’s Replication for a distributed environment also enables you to write code to convert one type of data in the mainframe and map it to data in the Oracle database. For example, your mainframe might be in London using European dates, yet your Oracle Server in New York needs to read in these dates and convert them. You can code functions that perform complex translations on columns.