Help Center/ GaussDB/ Centralized_8.x/ SQL Optimization/ Configuring Key Parameters for SQL Tuning
Updated on 2024-06-03 GMT+08:00

Configuring Key Parameters for SQL Tuning

This section describes key configuration parameters of the primary database node, which affect GaussDB SQL tuning. Contact the administrator to configure parameters.

Table 1 Parameters of the primary database node

Parameter/Reference Value

Description

enable_nestloop=on

Specifies how the optimizer uses Nest Loop Join. If this parameter is set to on, the optimizer preferentially uses Nest Loop Join. If it is set to off, the optimizer preferentially uses other methods, if any.

NOTE:

If you only want to temporarily change the value of this parameter during the current database connection (that is, the current session), execute the following SQL statement:

1
SET enable_nestloop to off;

By default, this parameter is set to on. Change the value as required. Generally, among the three join modes (Nested Loop, Merge Join, and Hash Join), Nested Loop is applicable to scenarios with small data volume or indexes, and Hash Join is applicable to big data analysis scenarios.

enable_bitmapscan=on

Specifies whether the optimizer uses bitmap scan. If the value is on, bitmap scan is used. If the value is off, it is not used.

NOTE:

If you only want to temporarily change the value of this parameter during the current database connection (that is, the current session), execute the following SQL statement:

1
SET enable_bitmapscan to off;

The bitmap scan applies only in the query condition where a > 1 and b > 1 and indexes are created on columns a and b. However, the performance of bitmapscan is sometimes inferior to that of indexscan. During tuning, if the query performance is poor and bitmapscan operators are in the execution plan, set this parameter to off and check whether the performance is improved.

enable_hashagg=on

Specifies whether the optimizer uses hash aggregate plans.

enable_hashjoin=on

Specifies whether the optimizer uses hash join plans.

enable_mergejoin=on

Specifies whether the optimizer uses merge join plans.

enable_indexscan=on

Specifies whether the optimizer uses index scan plans.

enable_indexonlyscan=on

Specifies whether the optimizer uses index-only scan plans.

enable_seqscan=on

Specifies whether the optimizer uses sequential scan plans. It is impossible to suppress sequential scans entirely, but setting this variable to off encourages the optimizer to choose other methods if available.

enable_sort=on

Specifies the optimizer sorting order. It is impossible to fully suppress explicit sorting, but setting this variable to off encourages the optimizer to choose other methods if available.

rewrite_rule

Specifies whether the optimizer enables the LAZYAGG, MAGICSET, PARTIALPUSH, DISABLEREP, UNIQUECHECK, INTARGETLIST, PREDPUSH, PREDPUSHFORCE, PREDPUSHNORMAL, DISABLE_PULLUP_EXPR_SUBLINK, ENABLE_SUBLINK_PULLUP_ENHANCED, DISABLE_PULLUP_NOT_IN_SUBLINK, DISABLE_ROWNUM_PUSHDOWN, or DISABLE_WINDOWAGG_PUSHDOWN rewriting rule.

sql_beta_feature

Specifies whether the optimizer enables the SEL_SEMI_POISSON, SEL_EXPR_INSTR, PARAM_PATH_GEN, RAND_COST_OPT, PARAM_PATH_OPT, PAGE_EST_OPT, CANONICAL_PATHKEY, PREDPUSH_SAME_LEVEL, PARTITION_FDW_ON, DISABLE_BITMAP_COST_WITH_LOSSY_PAGES, ENABLE_UPSERT_EXECUTE_GPLAN, or DISABLE_FASTPATH_INSERT beta feature.

var_eq_const_selectivity

Specifies whether the optimizer uses histograms to calculate the selectivity of integer constants.

partition_page_estimation

Specifies whether to optimize page estimation of partitioned table pages based on the pruning result. Only the partitioned table pages and local index pages are included. The global index pages are not included. The estimation formula is as follows:

Number of pages after estimation = Total number of pages in the partitioned table x (Number of partitions after pruning/Total number of partitions)

partition_iterator_elimination

Specifies whether to eliminate the partition iteration operators to improve execution efficiency when the partition pruning result of a partitioned table is a partition.

enable_functional_dependency

Specifies whether functional dependency statistics are used.

Setting this parameter to on will enable the following functions:

  1. The statistics about multiple columns generated by ANALYZE contain functional dependency statistics.
  2. Function dependency statistics are used to calculate the selectivity.

Setting this parameter to off will disable the following functions:

  1. The statistics about multiple columns generated by ANALYZE do not contain functional dependency statistics.
  2. Function dependency statistics are not used to calculate the selectivity.
NOTE:

The concept of functional dependency comes from the relational database normal form, which indicates the functional relationship between attributes. The concept of functional dependency statistics derives from the preceding concept. It indicates the ratio of the data volume that meets the functional relationship to the total data volume. Functional dependency statistics are a type of multi-column statistics, which can be used to improve the accuracy of selectivity estimation.

Functional dependency statistics are applicable to the where a = 1 and b = 1 condition, in which a and b must be attributes of the same table subject to equivalent constraints and the constraints are connected by AND. There are at least two constraints.

enable_seqscan_fusion

Specifies whether to enable the sequential scan noise floor elimination function.

enable_inner_unique_opt

Specifies whether the optimizer uses Inner Unique.