Choosing Physical Tuning Options › Data Compression

Data Compression

Conserving disk space

To conserve disk space, you can instruct the database to compress data before storage and decompress it after retrieval. There are three ways to compress and decompress data:

• CA IDMS Presspack
• IDMSCOMP and IDMSDCOM database procedures
• User-written procedure

These procedures are invoked automatically by the DBMS as data is stored and retrieved.

Note: Only CA IDMS Presspack is available for SQL-defined data.

Advantages and disadvantages of data compression

The following table summarizes the advantages and disadvantages of data compression.

Considerations for using CA IDMS Presspack

CA IDMS Presspack uses Huffman techniques to compress database entities. The techniques include assigning unique bit string codes of different lengths to single character and character strings. These codes substitute for the character and character strings in the entities.

To assign the codes, CA IDMS Presspack uses character and character-string frequencies of occurrence. It assigns shorter codes to the most frequently occurring characters and character strings. To those that occur less frequently, CA IDMS Presspack assigns longer codes.

CA IDMS Presspack compresses both textual and nontextual data.

For further information about CA IDMS Presspack, see CA IDMS Database Administration Guide and CA IDMS Presspack User Guide.

Considerations for using IDMSCOMP and IDMSDCOM

IDMSCOMP and IDMSDCOM are supplied with CA IDMS/DB. They are placed in the load (core-image) library at installation time and are also provided in source form so you can modify them if necessary. You can also write your own database procedure or use other commercially available compression/decompression procedures.

For further information about database procedures, see CA IDMS Database Administration Guide.

To compress data, IDMSCOMP performs the following conversion procedures:

• Converts repeating blanks into a 2-byte code.
• Converts repeating binary zeros into a 2-byte code.
• Converts other repeating characters into a 3-byte code.
• Converts any of a number of commonly used character pairs into a 1-byte code.

Data that does not fall into any of the above categories remains unchanged. Each group of unchanged data is prefixed by a 2-byte code. The following diagram shows the compression of contiguous blanks in an entity.

Considerations for user-written procedures

If writing your own compression procedures, you must follow conventions for writing database procedures.

For information on database procedures, see CA IDMS Database Administration Guide.

Guidelines for compression

Consider the following guidelines when deciding whether data should be compressed:

• When determining whether or not to compress/decompress an entity, you should consider whether the disk space saved justifies the CPU overhead incurred by the routines.
• The control portion of an entity is not compressible.

  The control portion of an entity includes all data elements up to the last key (CALC, sort, index). Since this portion of an entity is not compressible, it may mean that not enough compressible data exists to justify compression.

• Use compression/decompression procedures for entities that are not updated often. While the compression procedures save considerable disk space, it uses additional CPU time to perform its processing.
• Do not compress entities that start with large groups of repeating characters but lose them over time.
• IDMSCOMP/IDMSDCOM considerations:
  • IDMSCOMP and IDMSDCOM compression procedures operate most efficiently for entities whose occurrences usually contain sizable portions of blanks or binary zeros.
  • Don't use this compression for entities containing only small scattered groups of repeating characters.
  • Data that is stored in packed decimal format is not a good candidate for data compression.

Storage mode

If you decide to compress data in an entity, you should add a storage mode of C for the entity on the data structure diagram.