Parallel Data Loading

ttImportFromOracle utility provides two methods for loading data.

The two methods are:

• It generates a SQL script, LoadData.sql, consisting of a sequence of ttLoadFromOracle() calls. This script runs in ttIsql to load table data from Oracle into TimesTen. One disadvantage of this approach is that all data is loaded serially. There is one ttLoadFromOracle() call per table, and each call runs sequentially. As a result, this script is suitable for loading small volumes of data into a small number of tables, but it is relatively inefficient for large numbers of tables and/or large volumes of data. This script is used on both Windows and Unix/Linux platforms.
• When the target platform is Unix/Linux, an additional shell script, ttPDL.sh, is also generated (unless parallel data loading is explicitly disabled). This script loads individual tables in parallel (using multiple ttLoadFromOracle() calls per table) and also loads multiple tables concurrently. Various options are available to control the degree of parallelism, both globally and on a per-table basis. However, the default settings are chosen to provide good performance on Exalytics systems, so if you are using one, there is no need for fine-tuning. Parallel data loading using this script more efficiently utilizes the hardware resources of both the machine hosting TimesTen and the backend Oracle database server, significantly accelerating data loading and reducing overall load times.