Loading a Non-SQL Defined Database › Loading Database Records Using FASTLOAD › General Considerations

General Considerations

Always Load in Local Mode

You must load a database in local mode. Journaling is not required, and is not recommended when loading a database for these reasons:

• The load utility does not maintain checkpoints
• It's easier to re-run a failed job step than to recover the database
• Journaling can impact performance

Cross-Area Sets

• If the owner and member records of an automatic set exist in different database areas, load the areas together.
• If the owner and member records of a manual set exist in different database areas, either:
  • Load the areas together
  • Run a user-written program to connect the records after loading the entire database

CALC Records

The target page for CALC records to be loaded into a database can be determined in one of two ways:

• By the standard CA IDMS/DB CALC routine (IDMSCALC).
• By a user-written CALC exit routine (IDMSCLCX) that was compiled and link-edited with IDMSUXIT.

  Important: If you determine the target page using IDMSCALC, you must use it whenever the database is accessed; likewise, if you use the IDMSCLCX user exit, you must continue to use an IDMSUXIT module with which it is linked.

Note: For more information on enabling user exits by linking IDMSUXIT, refer to the "User Exits" section of the CA IDMS Systems Operations Guide.

Compressed Data

If the schema definition specifies compression for a record type, CA IDMS/DB compresses the record before it stores it during a load operation. Therefore, before you begin the load procedure, be sure the schema definition includes the information it requires to compress the record occurrences.

Reserving Space on the Page

To reserve space for the storage of additional records on a page or for increases in the length of records stored on a page, add an area override to the DMCL that specifies a page reserve. When the load is complete, you can remove the area override.

Buffers

The DMCL that you use to load the database should contain a local mode buffer that contains at least 10 pages. One large buffer should be sufficient. However, you may obtain performance improvements by assigning the files associated with each area to a separate buffer. If you don't have enough resources, then try to assign the files associated with the following areas to separate buffers:

• Index area
• Areas for which the owner record exists in one area and the member record exists in another area

Considerations for Large Databases

A large database should be loaded in portions. The FASTLOAD statement assumes that all record occurrences that are connected by automatic sets will be loaded at the same time. For a large database, this assumption can be limiting. To load a large database:

1. Group the record types so that there are not automatic sets between the groups
2. Load each group of record types
3. If manual set connections exist between records in different groups, connect the records by executing a user-written program

Subschema Requirements

The subschema that you use in the load process must:

• Include all records being loaded and all set relationships in which the records participate
• Allow all affected areas to be readied in exclusive update mode