Query Optimizer in MariaDB 10.4

Query Optimizer
in MariaDB 10.4
Sergei Petrunia,
Query Optimizer Developer
MariaDB Corporation
2019 MariaDB Developers Unconference
New York

New Optimizer features in MariaDB 10.4
● Optimizer trace
● Sampling for histogram collection
● Rowid filtering
● New default settings
● Condition Pushdown into IN-subqueries
● Condition Pushdown from HAVING into WHERE

New default settings for statistics
-histogram_size=0
+histogram_size=254
-histogram_type=SINGLE_PREC_HB
+histogram_type=SINGLE_PREC_HB
-optimizer_use_condition_selectivity=1
+optimizer_use_condition_selectivity=4
1 Use selectivity of predicates as in MariaDB 5.5.
2 Use selectivity of all range predicates supported by indexes.
3 Use selectivity of all range predicates estimated without histogram.
4 Use selectivity of all range predicates estimated with histogram.
● Do use condition selectivity
● Make use of EITS statistics (incl. Histograms if they are available)
-use_stat_tables=NEVER
+use_stat_tables=PREFERABLY_FOR_QUERIES
– But don’t collect stats unless explicitly told to do so
● Do build histograms when collecting EITS statistics

New default settings
-eq_range_index_dive_limit=10
+eq_range_index_dive_limit=200
● Join buffer will auto-size itself
-optimize_join_buffer_size=OFF
+optimize_join_buffer_size=ON
● Use index statistics (cardinality) instead of records_in_range for large IN-lists
– Just following MySQL here
– (can use ANALYZE for statements to see the size)

Histograms in MariaDB
● Introduced in MariaDB 10.0
– Manual command to collect, ANALYZE … PERSISTENT FOR …
– Optimizer settings to use them
– Histogram is collected from ALL table data
●
Other statistics (avg_frequency, avg_length), too.
● Results
– A few users
– Histogram collection is expensive
●
Cost of full table scan + full index scans, and even more than that

Histograms in MariaDB 10.4
● MariaDB 10.4
– “Bernoulli sampling” - roll the dice for each row
– Controlled with @@analyze_sample_percentage
●
100 (the default) – “use all data”
●
0 – (recommended) – “Determine sample ratio automatically”
● MySQL 8.0 also added histograms
– Uses Bernoulli sampling
– @@histogram_generation_max_mem_size=20MB.
● PostgreSQL has genuine random-jump sampling
– default_statistics_target

Histogram collection performance
● See MDEV-17886, (TODO: Vicentiu’s data?)
● Both MariaDB and MySQL: ANALYZE for columns is as fast as full table scan.
ANALYZE TABLE PERSISTENT FOR COLUMNS (...) INDEXES ();
● “Persistent for ALL” will also scan indexes
ANALYZE TABLE PERSISTENT FOR ALL;
● PostgreSQL is much faster with genuine sampling
– Vicentiu’s has a task in progress for this.

Histogram precision
● MariaDB histograms are very compact
– min/max column values, then 1-byte or 2-byte bounds (SINGLE|DOUBLE_PREC_HB)
– 255 bytes per histogram => 128 or 255 buckets max.
● MySQL
– Histogram is stored as JSON, bounds are stored as values
– 100 Buckets by default, max is 1024
●
In our tests, more buckets help in some cases
● PostgreSQL
– Histogram bounds stored as values
– 100 buckets by default, up to 10K allowed
● Testing is still in progress :-(, the obtained data varies.

Problem with correlated conditions
● Possible selectivities
– MIN(1/n, 1/m)
– (1/n) * (1/m)
– 0
select ...
from order_items
where shipdate='2015-12-15' AND item_name='christmas light'
'swimsuit'

Problem with correlated conditions
● PostgreSQL: Multi-variate statistics
– Detects functional dependencies, col1=F(col2)
– Only used for equality predicates
– Also #DISTINCT(a,b)
● MariaDB: MDEV-11107: Use table check constraints in optimizer
– Stalled?
select ...
from order_items
where shipdate='2015-12-15' AND item_name='christmas light'
'swimsuit'

Histograms: conclusions
● 10.4
– Sampling makes ANALYZE TABLE … PERSISTENT FOR COLUMNS
run at full-table-scan speed.
– @@analyze_sample_rows
● Further directions
– Do real sampling (in progress)
– More space for the histograms (?)
– Handle correlations (how?)

Optimizer trace
● Available in MySQL since MySQL 5.6
mysql> set optimizer_trace=1;
mysql> <query>;
mysql> select * from
-> information_schema.optimizer_trace;
{
"steps": [
{
"join_preparation": {
"select#": 1,
"steps": [
{
"expanded_query": "/* select#1 */ select `t1`.`col1` AS `col1`,`t1`.`col2`
AS `col2` from `t1` where (`t1`.`col1` < 4)"
}
]
}
},
{
"join_optimization": {
"select#": 1,
"steps": [
{
"condition_processing": {
"condition": "WHERE",
"original_condition": "(`t1`.`col1` < 4)",
"steps": [
{
"transformation": "equality_propagation",
"resulting_condition": "(`t1`.`col1` < 4)"
},
{
"transformation": "constant_propagation",
},
{
"transformation": "trivial_condition_removal",
}
]
}
},
{
● Now, similar feature in MariaDB

The goal is to understand the optimizer
● “Why was query plan X not chosen”
– Subquery was not converted into semi-join
●
This would exceed MAX_TABLES
– Subquery materialization was not used
●
Different collations
– Ref acess was not used
●
Incompatible collations
● What changed between the two hosts/versions
– diff trace_from_host1 trace_from_host2

Developer point of view
● The trace is always compiled in
● RAII-objects to start/end writing a trace
● Disabled trace added ~1-2% overhead
● Intend to add more tracing
– Expect to get more output

What is PK-filter: in details
SELECT *
FROM orders JOIN lineitem ON o_orderkey=l_orderkey
WHERE l_shipdate BETWEEN '1997-01-01' AND '1997-06-30' AND
o_totalprice between 200000 and 230000;
Filter for lineitem table built with condition
l_shipdate BETWEEN '1997-01-01' AND '1997-06-30':
is a container that contains primary keys of rows from lineitem which
l_shipdate value satisfy the above condition.

SELECT *
o_totalprice > 200000;
1. There is index i_l_shipdate on
lineitem(l_shipdate)

2.
SELECT *
o_totalprice between 200000 and 230000;

SELECT ...
FROM t1
WHERE (a < 2) AND
a IN
(
SELECT c
FROM t2
WHERE … AND (c < 2)
GROUP BY c
);
How condition pushdown is made
SELECT ...
FROM t1
WHERE (a < 2) AND
a IN
(
SELECT c
FROM t2
WHERE ...
GROUP BY c
);

Query Optimizer in MariaDB 10.4

More Related Content

What's hot (19)

Similar to Query Optimizer in MariaDB 10.4 (20)

More from Sergey Petrunya (16)

Recently uploaded (20)

Query Optimizer in MariaDB 10.4