Performing Processes in Parallel (SAP Library - Database Administration (Oracle) with SAPDBA)

Performing Processes in Parallel

You can speed up a reorganization by performing the following operations in parallel:

Data transport (ORACLE exp/imp, SAP unload/load, SQL*Loader)

The export (or import) is performed in parallel by setting the degree of parallelism (parallel option or profile parameter exp_imp_degree) >1 and distributing the tables to be exported on several different export processes. A corresponding number of export dump files is generated. The different export processes can be processed either sequentially or in parallel.

The time for the data export/import (with the ORACLE tools Export/Import) depends on the amount of data. These procedures take a large part of the total organization time within the reorganization process. You should perform the export/import in parallel for this reason.

Information can be found in the ORACLE documentation and in: Parallel Export/Import.

exp_imp_degree.

Creating objects (tables with CREATE TABLE AS SELECT, indices with CREATE INDEX)

The creation of large tables and their indices can take a very long time. You should, therefore, make these processes parallel. More information can be found in the ORACLE documentation.

SAPDBA supports the parallel creation of tables and indexes if you proceed as follows:

Set the degree of parallelism with the parameter index_degree in the profile init<DBSID>.dba (also for tables).

See index_degree.

Interactive entry of the degree of paralellity in the corresponding reorganization menu

Select the menu option Object handling ® Parallel object creation or (with CREATE TABLE AS SELECT) Data transfer ® Parallel table creation. SAPDBA displays the default value set in index_degree . Enter the required degree of parallelism.

It does not make sense to perform the specified processes in parallel if you select a degree of parallelism which is significantly larger than the number of existing CPUs (the degree of parallelism should not be larger than two or three times the number of CPUs).