Type of Linkage

Choosing Physical Tuning Options › Relationship Tuning Options › Linked Relationship Tuning Options › Type of Linkage

Type of Linkage

CA IDMS/DB supports the following types of linked relationships:

Chained—The DBMS maintains relationships based on internal information stored in the prefix of each entity occurrence. This information in the prefix contains the db-key of the logically next occurrence in the relationship.
Indexed—The DBMS maintains relationships through an index between a parent and related child occurrences. The bottom level of the index contains the db-keys of the related child occurrences. Each child occurrence contains an index pointer that points to the bottom level of the index.

Guidelines

As a general rule, use indexed for nonclustered relationships and chained for clustered relationships.

An indexed nonclustered relationship requires fewer I/Os to add or remove an entity occurrence than a chained nonclustered relationship. This is because the adjacent entity occurrences are not updated; only the index structure needs to be updated. In addition, fewer I/Os are required to retrieve a child occurrence by key in a nonclustered relationship if it is indexed rather than chained.

A chained relationship, on the other hand, requires less CPU overhead for maintenance and retrieval than an indexed relationship. It also requires less storage space because there is no index structure. For these reasons, it is a better choice than indexed for clustered relationships because I/Os are not generally a concern.

Note: For databases implemented with SQL, all linked clustered relationships are chained and all linked nonclustered relationships are indexed.

For further information on the structure of indexed relationships, see Chapter 15, "Determining the Size of the Database".

For further information on indexed relationships, see CA IDMS Database Administration Guide.

A comparison of indexed and chained relationships

The following table presents a comparison of indexed relationships and chained relationships.

Efficiency Considerations	Potential Impact
I/O	Indexed relationships often require fewer I/O operations to access child entities in nonclustered relationships, especially if the relationship is sorted.
CPU time	Chained relationships use less CPU time for processing of child entities than indexed relationships.
Space management	Chained relationships require less storage space than indexed relationships.
Contention	No difference.

Representing an indexed relationship

To represent an indexed relationship:

Name the relationship.
Specify whether the order is ASCending or DEScending for each key.
Identify the data element name(s) to be indexed.
Specify whether duplicate indexed items are allowed (blank) or not allowed (U for unique).
Specify whether the index key is to be compressed.

The following diagram shows the standard CA IDMS/DB notation for an indexed relationship. The index allows the DBMS to access all EXPERTISE occurrences associated with a particular skill based on skill level in descending order.

Shows the standard CA IDMS/DB notation for an indexed relationship.