Release 10.1A: OpenEdge Getting Started:
Database Essentials

Dumping and loading

Determining when to perform a dump and load is a constant struggle. If you set up your database correctly, your records should not get fragmented, and only your indexes should need reorganizing. Index reorganization is the primary benefit of a dump and load. Usually, about 80 to 90 percent of the benefit of a dump and load can be achieved with index maintenance via the index rebuild and compact utilities. However, you might need to dump and load to reorganize your database into different areas or extents due to changes in application requirements.

A fast dump and load process is important, since the system will not be available to users during the process. Before performing a dump and load, it is important to have a good backup of your existing database. This backup provides you with a fallback position should something go awry.

The basic dump and load options are as follows:

Data Dictionary dump and load
Bulk loader
Binary dump and load

Data Dictionary dump and load

This option has existed since the inception of Progress and is still viable today, provided you follow these rules:

Multi-thread both the dump and the load. Generally, you should add sessions on both the dump and load until you cause a bottleneck on the system.
Use all of the disks on the system evenly to achieve maximum throughput. For example, you might want to make use of your BI disks because they will be idle during the dump portion of the process.
Leave the indexes enabled during reload. This does not make for efficient indexes due to index splits during the load process, but since the indexes are built at the same time the data is loaded, you can take advantage of the multi-threaded nature of OpenEdge. The indexes can be reorganized later through the use of an index compress.

Bulk loader

This option is a good one because it is simple. The bulk loader files are loaded sequentially with the indexes turned off. This necessitates an index rebuild after the load. The bulk load process itself is fairly quick. It is not possible to run multiple copies of this utility simultaneously. Run your index rebuild at the end of processing all your bulk loads.

Binary dump and load

The binary dump and load is much faster than the previous methods described. It allows for multi-threading of both the dump and the load. It also supplies the option to build indexes during the load, eliminating the need for a separate second step to build indexes. This utility, when used with the index build option, provides the best overall dump and load performance in the vast majority of cases.