Administration Guide

Introduction

The AutoLoader is a tool that can:

1. TRANSFER data from one system (like MVS) to another system (like UNIX).
2. SPLIT (or partition) that data in parallel.
3. LOAD the data simultaneously on the corresponding database partitions.

The AutoLoader may be run in one of four modes:

1. SPLIT_AND_LOAD. Data is split (perhaps in parallel) and loaded simultaneously on the corresponding database partitions.
2. SPLIT_ONLY. Data is split (perhaps in parallel) and the output from the splitters is written to files in the SPLIT_FILE_LOCATION or in the AutoLoader current working directory.
3. LOAD_ONLY. Data is expected to be already split and contained in files with the following naming convention:

      filename.xxx
   

   where xxx is the node number. The split process is skipped and data is loaded simultaneously on the corresponding database partitions. The assumption is made that filename.xxx is found in the SPLIT_FILE_LOCATION or in the current AutoLoader working directory.
4. ANALYZE. An optimal partitioning map with even distribution across all database partitions is generated. It is recommended that a data file with a fairly large number of records from the largest table in the database be used for this analysis.

Before you continue, you should be familiar with db2split and the load process. Refer to Appendix N. "Splitting Data with db2split" and "Using the LOAD Utility" for more information on these topics.

[ DB2 List of Books | Search the DB2 Books ]