Indexing guidelines

Tivoli Netcool/OMNIbus supports hash and tree index structures. The hash index supports equality comparisons in SQL queries. The tree index is an ordered index that stores column values in a sorted structure, and allows a wider range of comparisons, including equality, in SQL queries. Consequently, a tree index can be used in range queries and in queries with an ORDER BY clause.

Indexes are rebuilt whenever the ObjectServer is restarted, and these indexes use up a small amount of memory rather than physical disk space.

You can create indexes on all ObjectServer tables except the tables in the system databases, such as the catalog and security databases. Tivoli Netcool/OMNIbus also supports the use of indexes by sub-select clauses in SQL queries.

Although there is no limit on the number of indexes that you can create on a table, you must use indexes sparingly. Indexes incur a performance overhead because they are updated when insert, update, or delete operations are performed on the table on which they are based. For tables such as alerts.status, which are updated frequently, creating a large number of indexes can adversely affect the overall performance of the ObjectServer. Evaluate the tradeoff between indexing for fast retrieval of data and the performance degradation during insert, update, and delete operations.

Avoid indexing tables that contain only small amounts of data. Also avoid indexing columns that contain data values which frequently change.

Columns that are defined as primary keys are, by default, uniquely indexed. These special, implicit indexes are not stored in the catalog.indexes table. The Serial column in the alerts.status table is indexed by default.

During SQL processing, both the restriction filter for the table and the WHERE clause in each SELECT, UPDATE, DELETE, FOR EACH ROW, and EVALUATE statement is examined to determine whether an index scan should be performed instead of a full table scan. An index scan is performed when one or more predicates fulfills the following conditions:

• The predicate uses the equality operator (=) in the format ColumnName = ConstantExpression, where ColumnName is an indexed column.
• The predicate uses the less than (<), less than or equal to (<=), greater than (>), or greater than or equal to (>=) operator, providing ColumnName is an indexed column of type tree.
• The predicate is not connected to another predicate by an OR operator. For example, if the Severity or Serial field is indexed, an index is not used in the following SQL query:

  select Summary from alerts.status where Severity > 3 or Serial = 102;