Teradata Optimization Settings

Topics:

Adapter optimization allows the RDBMS to perform the work for which it is best suited, reducing the volume of RDBMS-to-server communication and improving response time. It also enables the RDBMS to exploit its own internal optimization techniques.

For more information, see Optimizing Requests and Optimizing Requests to Pass Virtual Fields Defined as Constants.

Specifying the Block Size for Insert Processing

How to:

xThe Adapter for Teradata supports array insert if the release of Teradata Client is:

  • ODBC: TTU13.10 or higher.
  • CLI: TTU12.00 or higher.

This technique substantially reduces network traffic and CPU utilization.

Using high values increases the efficiency of requests involving many rows, at the cost of higher virtual storage requirements.

Tip: You can change this setting manually.

Syntax: How to Specify the Block Size for Insert Processing

In combination with LOADONLY, the block size for an INSERT applies to MODIFY INCLUDE requests. INSERTSIZE is also supported for parameterized DIRECT SQL INSERT statements.

ENGINE SQLDBC SET INSERTSIZE n

where:

SQLDBC

Indicates the adapter. You can omit this value if you previously issued the SET SQLENGINE command.

n

Is the number of rows to be inserted using array insert techniques. Accepted values are 1 to 5,000. The default value is 1. If the result set contains a column that has to be processed as a LOB, the INSERTSIZE value used for that result set is 1.

Improving Efficiency With Aggregate Awareness

x

Aggregate awareness substantially improves the efficiency of queries.

Note: For Teradata, aggregate awareness is handled internally. No adapter settings are required. However, the COLLECT STATISTICS command must be issued in Teradata.

Optimizing Non-Equality WHERE-Based Left Outer Joins

How to:

Reference:

A left outer join selects all records from the host table and matches them with records from the cross-referenced table. When no matching records exist, the host record is still retained, and default values (blank or zero) are assigned to the cross-referenced fields. The adapter can optimize any WHERE-based left outer join command in which the conditional expression is supported by the RDBMS.

Syntax: How to Specify a Conditional Left Outer JOIN

JOIN LEFT_OUTER FILE hostfile AT hfld1 [TAG tag1]
     [WITH hfld2]
     TO {UNIQUE|MULTIPLE} 
     FILE crfile AT crfld [TAG tag2] [AS joinname]
     [WHERE expression1;
     [WHERE expression2;
     ...]
 
END

where:

LEFT_OUTER

Specifies a left outer join. If you do not specify the type of join in the JOIN command, the ALL parameter setting determines the type of join to perform.

hostfile

Is the host Master File.

AT

Links the correct parent segment or host to the correct child or cross-referenced segment. The field values used as the AT parameter are not used to cause the link. They are used as segment references.

hfld1

Is the field name in the host Master File whose segment will be joined to the cross-referenced data source. The field name must be at the lowest level segment in its data source that is referenced.

tag1

Is the optional tag name that is used as a unique qualifier for fields and aliases in the host data source.

WITH hfld2

Is a data source field with which to associate a DEFINE-based conditional JOIN. For a DEFINE-based conditional join, the KEEPDEFINES setting must be ON, and you must create the virtual fields before issuing the JOIN command.

MULTIPLE

Specifies a one-to-many relationship between from_file and to_file. Note that ALL is a synonym for MULTIPLE.

UNIQUE

Specifies a one-to-one relationship between hostfile and crfile. Note that ONE is a synonym for UNIQUE.

Note: Unique returns only one instance and, if there is no matching instance in the cross-referenced file, it supplies default values (blank for alphanumeric fields and zero for numeric fields).

The unique join is a WebFOCUS concept. The RDBMS makes no distinction between unique and non-unique situations; it always retrieves all matching rows from the cross-referenced file.

If the RDBMS processes a join that the request specifies as unique, and if there are, in fact, multiple corresponding rows in the cross-referenced file, the RDBMS returns all matching rows. If, instead, optimization is disabled so that WebFOCUS processes the join, a different report results because WebFOCUS, respecting the unique join concept, returns only one cross-referenced row for each host row.

crfile

Is the cross-referenced Master File.

crfld

Is the join field name in the cross-referenced Master File. It can be any field in the segment.

tag2

Is the optional tag name that is used as a unique qualifier for fields and aliases in the cross-referenced data source.

joinname

Is the name associated with the joined structure.

expression1, expression2

Are any expressions that are acceptable in a DEFINE FILE command. All fields used in the expressions must lie on a single path.

Reference: Conditions for WHERE-Based Outer Join Optimization

  • In order for a WHERE-based left outer join to be optimized, the expressions must be optimizable for the RDBMS involved and at least one of the following conditions must be true:
    • The JOIN WHERE command contains at least one field1 EQ field2 predicate in which field1 is in table1 and field2 is in table2.

      or

    • The right table has a key or a unique index that does not contain NULL data.

      or

    • The right table contains at least one "NOT NULL" column that does not have a long data type (such as TEXT or IMAGE).
  • The adapter SQLJOIN OUTER setting must be ON (the default).

Example: Optimizing a Non-Equality Left-Outer Join

The following request creates a left outer conditional join between two Teradata data sources and reports against the joined data sources. The STMTRACE is turned on in order to view the SQL generated for this request:

SET TRACEUSER = ON
SET TRACEOFF = ALL
SET TRACEON = STMTRACE//CLIENT
JOIN LEFT_OUTER FILE baseapp/EQUIP AT CARS
TO ALL FILE baseapp/CARREC AT CARC
WHERE CARS NE CARC;
END
TABLE FILE baseapp/EQUIP
PRINT CARS CARC STANDARD
BY MODEL
END

The WebFOCUS request is translated to a single Teradata SELECT statement that incorporates the left outer join, and the non-equality condition is passed to the RDBMS in the ON clause:

SELECT T1."CARS"(CHAR( 16)),T1."STANDARD"(CHAR( 40)),           
T2."CARC"(CHAR( 16)),T2."MODEL"(CHAR( 24)) FROM ( EQUIP T1 LEFT  
OUTER JOIN CARREC T2 ON (T1."CARS" <> T2."CARC") ) ORDER BY      
T2."MODEL"; 

WebFOCUS

Feedback