Enables optimization for aggregations on dictionaries. This can also be specified
on a per-query basis using the
dictionary_aggregation session property.
Compute hash codes for distribution, joins, and aggregations early during execution,
allowing result to be shared between operations later in the query. This can reduce
CPU usage by avoiding computing the same hash multiple times, but at the cost of
additional network transfer for the hashes. In most cases it decreases overall
query processing time. This can also be specified on a per-query basis using the
optimize_hash_generation session property.
It is often helpful to disable this property, when using EXPLAIN in order to make the query plan easier to read.
Enable optimization of some aggregations by using values that are stored as metadata.
This allows Trino to execute some simple queries in constant time. Currently, this
optimization applies to
approx_distinct of partition
keys and other aggregation insensitive to the cardinality of the input,including
DISTINCT aggregates. Using this may speed up some queries significantly.
The main drawback is that it can produce incorrect results, if the connector returns partition keys for partitions that have no rows. In particular, the Hive connector can return empty partitions, if they were created by other systems. Trino cannot create them.
When an aggregation is above an outer join and all columns from the outer side of the join are in the grouping clause, the aggregation is pushed below the outer join. This optimization is particularly useful for correlated scalar subqueries, which get rewritten to an aggregation over an outer join. For example:
SELECT * FROM item i WHERE i.i_current_price > ( SELECT AVG(j.i_current_price) FROM item j WHERE i.i_category = j.i_category);
Enabling this optimization can substantially speed up queries by reducing
the amount of data that needs to be processed by the join. However, it may slow down some
queries that have very selective joins. This can also be specified on a per-query basis using
push_aggregation_through_join session property.
Parallelize writes when using
UNION ALL in queries that write data. This improves the
speed of writing output tables in
UNION ALL queries, because these writes do not require
additional synchronization when collecting results. Enabling this optimization can improve
UNION ALL speed, when write speed is not yet saturated. However, it may slow down queries
in an already heavily loaded system. This can also be specified on a per-query basis
push_table_write_through_union session property.
The join reordering strategy to use.
NONE maintains the order the tables are listed in the
ELIMINATE_CROSS_JOINS reorders joins to eliminate cross joins, where possible, and
otherwise maintains the original query order. When reordering joins, it also strives to maintain the
original table order as much as possible.
AUTOMATIC enumerates possible orders, and uses
statistics-based cost estimation to determine the least cost order. If stats are not available, or if
for any reason a cost could not be computed, the
ELIMINATE_CROSS_JOINS strategy is used. This can
be specified on a per-query basis using the
join_reordering_strategy session property.
When optimizer.join-reordering-strategy is set to cost-based, this property determines the maximum number of joins that can be reordered at once.
The number of possible join orders scales factorially with the number of relations, so increasing this value can cause serious performance issues.
Reduces number of rows produced by joins when optimizer detects that duplicated join output rows can be skipped.
Use connector provided table node partitioning when reading tables.
For example, table node partitioning corresponds to Hive table buckets.
When set to
true and minimal partition to task ratio is matched or exceeded,
each table partition is read by a separate worker. The minimal ratio is defined in
Partition reader assignments are distributed across workers for parallel processing. Use of table scan node partitioning can improve query performance by reducing query complexity. For example, cluster wide data reshuffling might not be needed when processing an aggregation query. However, query parallelism might be reduced when partition count is low compared to number of workers.
Specifies minimal bucket to task ratio that has to be matched or exceeded in order to use table scan node partitioning. When the table bucket count is small compared to the number of workers, then the table scan is distributed across all workers for improved parallelism.