Google Git
Sign in
apache/cloudberry/refs/heads/string_replace_fix/./doc/src/sgml/perform.sgml
blob: 749d469374467c157ad3b83e70afd6a641b0f2e4 [file]
<!-- doc/src/sgml/perform.sgml -->
<chapter id="performance-tips">
<title>Performance Tips</title>
<indexterm zone="performance-tips">
<primary>performance</primary>
</indexterm>
<para>
Query performance can be affected by many things. Some of these can
be controlled by the user, while others are fundamental to the underlying
design of the system. This chapter provides some hints about understanding
and tuning <productname>PostgreSQL</productname> performance.
</para>
<sect1 id="using-explain">
<title>Using <command>EXPLAIN</command></title>
<indexterm zone="using-explain">
<primary>EXPLAIN</primary>
</indexterm>
<indexterm zone="using-explain">
<primary>query plan</primary>
</indexterm>
<para>
<productname>PostgreSQL</productname> devises a <firstterm>query
plan</firstterm> for each query it receives. Choosing the right
plan to match the query structure and the properties of the data
is absolutely critical for good performance, so the system includes
a complex <firstterm>planner</firstterm> that tries to choose good plans.
You can use the <link linkend="sql-explain"><command>EXPLAIN</command></link> command
to see what query plan the planner creates for any query.
Plan-reading is an art that requires some experience to master,
but this section attempts to cover the basics.
</para>
<para>
Examples in this section are drawn from the regression test database
after doing a <command>VACUUM ANALYZE</command>, using 9.3 development sources.
You should be able to get similar results if you try the examples
yourself, but your estimated costs and row counts might vary slightly
because <command>ANALYZE</command>'s statistics are random samples rather
than exact, and because costs are inherently somewhat platform-dependent.
</para>
<para>
The examples use <command>EXPLAIN</command>'s default <quote>text</quote> output
format, which is compact and convenient for humans to read.
If you want to feed <command>EXPLAIN</command>'s output to a program for further
analysis, you should use one of its machine-readable output formats
(XML, JSON, or YAML) instead.
</para>
<sect2 id="using-explain-basics">
<title><command>EXPLAIN</command> Basics</title>
<para>
The structure of a query plan is a tree of <firstterm>plan nodes</firstterm>.
Nodes at the bottom level of the tree are scan nodes: they return raw rows
from a table. There are different types of scan nodes for different
table access methods: sequential scans, index scans, and bitmap index
scans. There are also non-table row sources, such as <literal>VALUES</literal>
clauses and set-returning functions in <literal>FROM</literal>, which have their
own scan node types.
If the query requires joining, aggregation, sorting, or other
operations on the raw rows, then there will be additional nodes
above the scan nodes to perform these operations. Again,
there is usually more than one possible way to do these operations,
so different node types can appear here too. The output
of <command>EXPLAIN</command> has one line for each node in the plan
tree, showing the basic node type plus the cost estimates that the planner
made for the execution of that plan node. Additional lines might appear,
indented from the node's summary line,
to show additional properties of the node.
The very first line (the summary line for the topmost
node) has the estimated total execution cost for the plan; it is this
number that the planner seeks to minimize.
</para>
<para>
Here is a trivial example, just to show what the output looks like:
<screen>
EXPLAIN SELECT * FROM tenk1;
QUERY PLAN
-------------------------------------------------------------
Seq Scan on tenk1 (cost=0.00..458.00 rows=10000 width=244)
</screen>
</para>
<para>
Since this query has no <literal>WHERE</literal> clause, it must scan all the
rows of the table, so the planner has chosen to use a simple sequential
scan plan. The numbers that are quoted in parentheses are (left
to right):
<itemizedlist>
<listitem>
<para>
Estimated start-up cost. This is the time expended before the output
phase can begin, e.g., time to do the sorting in a sort node.
</para>
</listitem>
<listitem>
<para>
Estimated total cost. This is stated on the assumption that the plan
node is run to completion, i.e., all available rows are retrieved.
In practice a node's parent node might stop short of reading all
available rows (see the <literal>LIMIT</literal> example below).
</para>
</listitem>
<listitem>
<para>
Estimated number of rows output by this plan node. Again, the node
is assumed to be run to completion.
</para>
</listitem>
<listitem>
<para>
Estimated average width of rows output by this plan node (in bytes).
</para>
</listitem>
</itemizedlist>
</para>
<para>
The costs are measured in arbitrary units determined by the planner's
cost parameters (see <xref linkend="runtime-config-query-constants"/>).
Traditional practice is to measure the costs in units of disk page
fetches; that is, <xref linkend="guc-seq-page-cost"/> is conventionally
set to <literal>1.0</literal> and the other cost parameters are set relative
to that. The examples in this section are run with the default cost
parameters.
</para>
<para>
It's important to understand that the cost of an upper-level node includes
the cost of all its child nodes. It's also important to realize that
the cost only reflects things that the planner cares about.
In particular, the cost does not consider the time spent transmitting
result rows to the client, which could be an important
factor in the real elapsed time; but the planner ignores it because
it cannot change it by altering the plan. (Every correct plan will
output the same row set, we trust.)
</para>
<para>
The <literal>rows</literal> value is a little tricky because it is
not the number of rows processed or scanned by the
plan node, but rather the number emitted by the node. This is often
less than the number scanned, as a result of filtering by any
<literal>WHERE</literal>-clause conditions that are being applied at the node.
Ideally the top-level rows estimate will approximate the number of rows
actually returned, updated, or deleted by the query.
</para>
<para>
Returning to our example:
<screen>
EXPLAIN SELECT * FROM tenk1;
QUERY PLAN
-------------------------------------------------------------
Seq Scan on tenk1 (cost=0.00..458.00 rows=10000 width=244)
</screen>
</para>
<para>
These numbers are derived very straightforwardly. If you do:
<programlisting>
SELECT relpages, reltuples FROM pg_class WHERE relname = 'tenk1';
</programlisting>
you will find that <classname>tenk1</classname> has 358 disk
pages and 10000 rows. The estimated cost is computed as (disk pages read *
<xref linkend="guc-seq-page-cost"/>) + (rows scanned *
<xref linkend="guc-cpu-tuple-cost"/>). By default,
<varname>seq_page_cost</varname> is 1.0 and <varname>cpu_tuple_cost</varname> is 0.01,
so the estimated cost is (358 * 1.0) + (10000 * 0.01) = 458.
</para>
<para>
Now let's modify the query to add a <literal>WHERE</literal> condition:
<screen>
EXPLAIN SELECT * FROM tenk1 WHERE unique1 &lt; 7000;
QUERY PLAN
------------------------------------------------------------
Seq Scan on tenk1 (cost=0.00..483.00 rows=7001 width=244)
Filter: (unique1 &lt; 7000)
</screen>
Notice that the <command>EXPLAIN</command> output shows the <literal>WHERE</literal>
clause being applied as a <quote>filter</quote> condition attached to the Seq
Scan plan node. This means that
the plan node checks the condition for each row it scans, and outputs
only the ones that pass the condition.
The estimate of output rows has been reduced because of the
<literal>WHERE</literal> clause.
However, the scan will still have to visit all 10000 rows, so the cost
hasn't decreased; in fact it has gone up a bit (by 10000 * <xref
linkend="guc-cpu-operator-cost"/>, to be exact) to reflect the extra CPU
time spent checking the <literal>WHERE</literal> condition.
</para>
<para>
The actual number of rows this query would select is 7000, but the <literal>rows</literal>
estimate is only approximate. If you try to duplicate this experiment,
you will probably get a slightly different estimate; moreover, it can
change after each <command>ANALYZE</command> command, because the
statistics produced by <command>ANALYZE</command> are taken from a
randomized sample of the table.
</para>
<para>
Now, let's make the condition more restrictive:
<screen>
EXPLAIN SELECT * FROM tenk1 WHERE unique1 &lt; 100;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------
Bitmap Heap Scan on tenk1 (cost=5.07..229.20 rows=101 width=244)
Recheck Cond: (unique1 &lt; 100)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..5.04 rows=101 width=0)
Index Cond: (unique1 &lt; 100)
</screen>
Here the planner has decided to use a two-step plan: the child plan
node visits an index to find the locations of rows matching the index
condition, and then the upper plan node actually fetches those rows
from the table itself. Fetching rows separately is much more
expensive than reading them sequentially, but because not all the pages
of the table have to be visited, this is still cheaper than a sequential
scan. (The reason for using two plan levels is that the upper plan
node sorts the row locations identified by the index into physical order
before reading them, to minimize the cost of separate fetches.
The <quote>bitmap</quote> mentioned in the node names is the mechanism that
does the sorting.)
</para>
<para>
Now let's add another condition to the <literal>WHERE</literal> clause:
<screen>
EXPLAIN SELECT * FROM tenk1 WHERE unique1 &lt; 100 AND stringu1 = 'xxx';
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------
Bitmap Heap Scan on tenk1 (cost=5.04..229.43 rows=1 width=244)
Recheck Cond: (unique1 &lt; 100)
Filter: (stringu1 = 'xxx'::name)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..5.04 rows=101 width=0)
Index Cond: (unique1 &lt; 100)
</screen>
The added condition <literal>stringu1 = 'xxx'</literal> reduces the
output row count estimate, but not the cost because we still have to visit
the same set of rows. Notice that the <literal>stringu1</literal> clause
cannot be applied as an index condition, since this index is only on
the <literal>unique1</literal> column. Instead it is applied as a filter on
the rows retrieved by the index. Thus the cost has actually gone up
slightly to reflect this extra checking.
</para>
<para>
In some cases the planner will prefer a <quote>simple</quote> index scan plan:
<screen>
EXPLAIN SELECT * FROM tenk1 WHERE unique1 = 42;
QUERY PLAN
-------------------------------------------------------------------&zwsp;----------
Index Scan using tenk1_unique1 on tenk1 (cost=0.29..8.30 rows=1 width=244)
Index Cond: (unique1 = 42)
</screen>
In this type of plan the table rows are fetched in index order, which
makes them even more expensive to read, but there are so few that the
extra cost of sorting the row locations is not worth it. You'll most
often see this plan type for queries that fetch just a single row. It's
also often used for queries that have an <literal>ORDER BY</literal> condition
that matches the index order, because then no extra sorting step is needed
to satisfy the <literal>ORDER BY</literal>. In this example, adding
<literal>ORDER BY unique1</literal> would use the same plan because the
index already implicitly provides the requested ordering.
</para>
<para>
The planner may implement an <literal>ORDER BY</literal> clause in several
ways. The above example shows that such an ordering clause may be
implemented implicitly. The planner may also add an explicit
<literal>sort</literal> step:
<screen>
EXPLAIN SELECT * FROM tenk1 ORDER BY unique1;
QUERY PLAN
-------------------------------------------------------------------
Sort (cost=1109.39..1134.39 rows=10000 width=244)
Sort Key: unique1
-> Seq Scan on tenk1 (cost=0.00..445.00 rows=10000 width=244)
</screen>
If a part of the plan guarantees an ordering on a prefix of the
required sort keys, then the planner may instead decide to use an
<literal>incremental sort</literal> step:
<screen>
EXPLAIN SELECT * FROM tenk1 ORDER BY four, ten LIMIT 100;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------------------------------
Limit (cost=521.06..538.05 rows=100 width=244)
-> Incremental Sort (cost=521.06..2220.95 rows=10000 width=244)
Sort Key: four, ten
Presorted Key: four
-> Index Scan using index_tenk1_on_four on tenk1 (cost=0.29..1510.08 rows=10000 width=244)
</screen>
Compared to regular sorts, sorting incrementally allows returning tuples
before the entire result set has been sorted, which particularly enables
optimizations with <literal>LIMIT</literal> queries. It may also reduce
memory usage and the likelihood of spilling sorts to disk, but it comes at
the cost of the increased overhead of splitting the result set into multiple
sorting batches.
</para>
<para>
If there are separate indexes on several of the columns referenced
in <literal>WHERE</literal>, the planner might choose to use an AND or OR
combination of the indexes:
<screen>
EXPLAIN SELECT * FROM tenk1 WHERE unique1 &lt; 100 AND unique2 &gt; 9000;
QUERY PLAN
-------------------------------------------------------------------&zwsp;------------------
Bitmap Heap Scan on tenk1 (cost=25.08..60.21 rows=10 width=244)
Recheck Cond: ((unique1 &lt; 100) AND (unique2 &gt; 9000))
-&gt; BitmapAnd (cost=25.08..25.08 rows=10 width=0)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..5.04 rows=101 width=0)
Index Cond: (unique1 &lt; 100)
-&gt; Bitmap Index Scan on tenk1_unique2 (cost=0.00..19.78 rows=999 width=0)
Index Cond: (unique2 &gt; 9000)
</screen>
But this requires visiting both indexes, so it's not necessarily a win
compared to using just one index and treating the other condition as
a filter. If you vary the ranges involved you'll see the plan change
accordingly.
</para>
<para>
Here is an example showing the effects of <literal>LIMIT</literal>:
<screen>
EXPLAIN SELECT * FROM tenk1 WHERE unique1 &lt; 100 AND unique2 &gt; 9000 LIMIT 2;
QUERY PLAN
-------------------------------------------------------------------&zwsp;------------------
Limit (cost=0.29..14.48 rows=2 width=244)
-&gt; Index Scan using tenk1_unique2 on tenk1 (cost=0.29..71.27 rows=10 width=244)
Index Cond: (unique2 &gt; 9000)
Filter: (unique1 &lt; 100)
</screen>
</para>
<para>
This is the same query as above, but we added a <literal>LIMIT</literal> so that
not all the rows need be retrieved, and the planner changed its mind about
what to do. Notice that the total cost and row count of the Index Scan
node are shown as if it were run to completion. However, the Limit node
is expected to stop after retrieving only a fifth of those rows, so its
total cost is only a fifth as much, and that's the actual estimated cost
of the query. This plan is preferred over adding a Limit node to the
previous plan because the Limit could not avoid paying the startup cost
of the bitmap scan, so the total cost would be something over 25 units
with that approach.
</para>
<para>
Let's try joining two tables, using the columns we have been discussing:
<screen>
EXPLAIN SELECT *
FROM tenk1 t1, tenk2 t2
WHERE t1.unique1 &lt; 10 AND t1.unique2 = t2.unique2;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-------------------
Nested Loop (cost=4.65..118.62 rows=10 width=488)
-&gt; Bitmap Heap Scan on tenk1 t1 (cost=4.36..39.47 rows=10 width=244)
Recheck Cond: (unique1 &lt; 10)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..4.36 rows=10 width=0)
Index Cond: (unique1 &lt; 10)
-&gt; Index Scan using tenk2_unique2 on tenk2 t2 (cost=0.29..7.91 rows=1 width=244)
Index Cond: (unique2 = t1.unique2)
</screen>
</para>
<para>
In this plan, we have a nested-loop join node with two table scans as
inputs, or children. The indentation of the node summary lines reflects
the plan tree structure. The join's first, or <quote>outer</quote>, child
is a bitmap scan similar to those we saw before. Its cost and row count
are the same as we'd get from <literal>SELECT ... WHERE unique1 &lt; 10</literal>
because we are
applying the <literal>WHERE</literal> clause <literal>unique1 &lt; 10</literal>
at that node.
The <literal>t1.unique2 = t2.unique2</literal> clause is not relevant yet,
so it doesn't affect the row count of the outer scan. The nested-loop
join node will run its second,
or <quote>inner</quote> child once for each row obtained from the outer child.
Column values from the current outer row can be plugged into the inner
scan; here, the <literal>t1.unique2</literal> value from the outer row is available,
so we get a plan and costs similar to what we saw above for a simple
<literal>SELECT ... WHERE t2.unique2 = <replaceable>constant</replaceable></literal> case.
(The estimated cost is actually a bit lower than what was seen above,
as a result of caching that's expected to occur during the repeated
index scans on <literal>t2</literal>.) The
costs of the loop node are then set on the basis of the cost of the outer
scan, plus one repetition of the inner scan for each outer row (10 * 7.91,
here), plus a little CPU time for join processing.
</para>
<para>
In this example the join's output row count is the same as the product
of the two scans' row counts, but that's not true in all cases because
there can be additional <literal>WHERE</literal> clauses that mention both tables
and so can only be applied at the join point, not to either input scan.
Here's an example:
<screen>
EXPLAIN SELECT *
FROM tenk1 t1, tenk2 t2
WHERE t1.unique1 &lt; 10 AND t2.unique2 &lt; 10 AND t1.hundred &lt; t2.hundred;
QUERY PLAN
-------------------------------------------------------------------&zwsp;--------------------------
Nested Loop (cost=4.65..49.46 rows=33 width=488)
Join Filter: (t1.hundred &lt; t2.hundred)
-&gt; Bitmap Heap Scan on tenk1 t1 (cost=4.36..39.47 rows=10 width=244)
Recheck Cond: (unique1 &lt; 10)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..4.36 rows=10 width=0)
Index Cond: (unique1 &lt; 10)
-&gt; Materialize (cost=0.29..8.51 rows=10 width=244)
-&gt; Index Scan using tenk2_unique2 on tenk2 t2 (cost=0.29..8.46 rows=10 width=244)
Index Cond: (unique2 &lt; 10)
</screen>
The condition <literal>t1.hundred &lt; t2.hundred</literal> can't be
tested in the <literal>tenk2_unique2</literal> index, so it's applied at the
join node. This reduces the estimated output row count of the join node,
but does not change either input scan.
</para>
<para>
Notice that here the planner has chosen to <quote>materialize</quote> the inner
relation of the join, by putting a Materialize plan node atop it. This
means that the <literal>t2</literal> index scan will be done just once, even
though the nested-loop join node needs to read that data ten times, once
for each row from the outer relation. The Materialize node saves the data
in memory as it's read, and then returns the data from memory on each
subsequent pass.
</para>
<para>
When dealing with outer joins, you might see join plan nodes with both
<quote>Join Filter</quote> and plain <quote>Filter</quote> conditions attached.
Join Filter conditions come from the outer join's <literal>ON</literal> clause,
so a row that fails the Join Filter condition could still get emitted as
a null-extended row. But a plain Filter condition is applied after the
outer-join rules and so acts to remove rows unconditionally. In an inner
join there is no semantic difference between these types of filters.
</para>
<para>
If we change the query's selectivity a bit, we might get a very different
join plan:
<screen>
EXPLAIN SELECT *
FROM tenk1 t1, tenk2 t2
WHERE t1.unique1 &lt; 100 AND t1.unique2 = t2.unique2;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------------------
Hash Join (cost=230.47..713.98 rows=101 width=488)
Hash Cond: (t2.unique2 = t1.unique2)
-&gt; Seq Scan on tenk2 t2 (cost=0.00..445.00 rows=10000 width=244)
-&gt; Hash (cost=229.20..229.20 rows=101 width=244)
-&gt; Bitmap Heap Scan on tenk1 t1 (cost=5.07..229.20 rows=101 width=244)
Recheck Cond: (unique1 &lt; 100)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..5.04 rows=101 width=0)
Index Cond: (unique1 &lt; 100)
</screen>
</para>
<para>
Here, the planner has chosen to use a hash join, in which rows of one
table are entered into an in-memory hash table, after which the other
table is scanned and the hash table is probed for matches to each row.
Again note how the indentation reflects the plan structure: the bitmap
scan on <literal>tenk1</literal> is the input to the Hash node, which constructs
the hash table. That's then returned to the Hash Join node, which reads
rows from its outer child plan and searches the hash table for each one.
</para>
<para>
Another possible type of join is a merge join, illustrated here:
<screen>
EXPLAIN SELECT *
FROM tenk1 t1, onek t2
WHERE t1.unique1 &lt; 100 AND t1.unique2 = t2.unique2;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------------------
Merge Join (cost=198.11..268.19 rows=10 width=488)
Merge Cond: (t1.unique2 = t2.unique2)
-&gt; Index Scan using tenk1_unique2 on tenk1 t1 (cost=0.29..656.28 rows=101 width=244)
Filter: (unique1 &lt; 100)
-&gt; Sort (cost=197.83..200.33 rows=1000 width=244)
Sort Key: t2.unique2
-&gt; Seq Scan on onek t2 (cost=0.00..148.00 rows=1000 width=244)
</screen>
</para>
<para>
Merge join requires its input data to be sorted on the join keys. In this
plan the <literal>tenk1</literal> data is sorted by using an index scan to visit
the rows in the correct order, but a sequential scan and sort is preferred
for <literal>onek</literal>, because there are many more rows to be visited in
that table.
(Sequential-scan-and-sort frequently beats an index scan for sorting many rows,
because of the nonsequential disk access required by the index scan.)
</para>
<para>
One way to look at variant plans is to force the planner to disregard
whatever strategy it thought was the cheapest, using the enable/disable
flags described in <xref linkend="runtime-config-query-enable"/>.
(This is a crude tool, but useful. See
also <xref linkend="explicit-joins"/>.)
For example, if we're unconvinced that sequential-scan-and-sort is the best way to
deal with table <literal>onek</literal> in the previous example, we could try
<screen>
SET enable_sort = off;
EXPLAIN SELECT *
FROM tenk1 t1, onek t2
WHERE t1.unique1 &lt; 100 AND t1.unique2 = t2.unique2;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------------------
Merge Join (cost=0.56..292.65 rows=10 width=488)
Merge Cond: (t1.unique2 = t2.unique2)
-&gt; Index Scan using tenk1_unique2 on tenk1 t1 (cost=0.29..656.28 rows=101 width=244)
Filter: (unique1 &lt; 100)
-&gt; Index Scan using onek_unique2 on onek t2 (cost=0.28..224.79 rows=1000 width=244)
</screen>
which shows that the planner thinks that sorting <literal>onek</literal> by
index-scanning is about 12% more expensive than sequential-scan-and-sort.
Of course, the next question is whether it's right about that.
We can investigate that using <command>EXPLAIN ANALYZE</command>, as discussed
below.
</para>
</sect2>
<sect2 id="using-explain-analyze">
<title><command>EXPLAIN ANALYZE</command></title>
<para>
It is possible to check the accuracy of the planner's estimates
by using <command>EXPLAIN</command>'s <literal>ANALYZE</literal> option. With this
option, <command>EXPLAIN</command> actually executes the query, and then displays
the true row counts and true run time accumulated within each plan node,
along with the same estimates that a plain <command>EXPLAIN</command>
shows. For example, we might get a result like this:
<screen>
EXPLAIN ANALYZE SELECT *
FROM tenk1 t1, tenk2 t2
WHERE t1.unique1 &lt; 10 AND t1.unique2 = t2.unique2;
QUERY PLAN
-------------------------------------------------------------------&zwsp;--------------------------------------------------------------
Nested Loop (cost=4.65..118.62 rows=10 width=488) (actual time=0.128..0.377 rows=10 loops=1)
-&gt; Bitmap Heap Scan on tenk1 t1 (cost=4.36..39.47 rows=10 width=244) (actual time=0.057..0.121 rows=10 loops=1)
Recheck Cond: (unique1 &lt; 10)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..4.36 rows=10 width=0) (actual time=0.024..0.024 rows=10 loops=1)
Index Cond: (unique1 &lt; 10)
-&gt; Index Scan using tenk2_unique2 on tenk2 t2 (cost=0.29..7.91 rows=1 width=244) (actual time=0.021..0.022 rows=1 loops=10)
Index Cond: (unique2 = t1.unique2)
Planning time: 0.181 ms
Execution time: 0.501 ms
</screen>
Note that the <quote>actual time</quote> values are in milliseconds of
real time, whereas the <literal>cost</literal> estimates are expressed in
arbitrary units; so they are unlikely to match up.
The thing that's usually most important to look for is whether the
estimated row counts are reasonably close to reality. In this example
the estimates were all dead-on, but that's quite unusual in practice.
</para>
<para>
In some query plans, it is possible for a subplan node to be executed more
than once. For example, the inner index scan will be executed once per
outer row in the above nested-loop plan. In such cases, the
<literal>loops</literal> value reports the
total number of executions of the node, and the actual time and rows
values shown are averages per-execution. This is done to make the numbers
comparable with the way that the cost estimates are shown. Multiply by
the <literal>loops</literal> value to get the total time actually spent in
the node. In the above example, we spent a total of 0.220 milliseconds
executing the index scans on <literal>tenk2</literal>.
</para>
<para>
In some cases <command>EXPLAIN ANALYZE</command> shows additional execution
statistics beyond the plan node execution times and row counts.
For example, Sort and Hash nodes provide extra information:
<screen>
EXPLAIN ANALYZE SELECT *
FROM tenk1 t1, tenk2 t2
WHERE t1.unique1 &lt; 100 AND t1.unique2 = t2.unique2 ORDER BY t1.fivethous;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-------------------------------------------------------------------&zwsp;------
Sort (cost=717.34..717.59 rows=101 width=488) (actual time=7.761..7.774 rows=100 loops=1)
Sort Key: t1.fivethous
Sort Method: quicksort Memory: 77kB
-&gt; Hash Join (cost=230.47..713.98 rows=101 width=488) (actual time=0.711..7.427 rows=100 loops=1)
Hash Cond: (t2.unique2 = t1.unique2)
-&gt; Seq Scan on tenk2 t2 (cost=0.00..445.00 rows=10000 width=244) (actual time=0.007..2.583 rows=10000 loops=1)
-&gt; Hash (cost=229.20..229.20 rows=101 width=244) (actual time=0.659..0.659 rows=100 loops=1)
Buckets: 1024 Batches: 1 Memory Usage: 28kB
-&gt; Bitmap Heap Scan on tenk1 t1 (cost=5.07..229.20 rows=101 width=244) (actual time=0.080..0.526 rows=100 loops=1)
Recheck Cond: (unique1 &lt; 100)
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..5.04 rows=101 width=0) (actual time=0.049..0.049 rows=100 loops=1)
Index Cond: (unique1 &lt; 100)
Planning time: 0.194 ms
Execution time: 8.008 ms
</screen>
The Sort node shows the sort method used (in particular, whether the sort
was in-memory or on-disk) and the amount of memory or disk space needed.
The Hash node shows the number of hash buckets and batches as well as the
peak amount of memory used for the hash table. (If the number of batches
exceeds one, there will also be disk space usage involved, but that is not
shown.)
</para>
<para>
Another type of extra information is the number of rows removed by a
filter condition:
<screen>
EXPLAIN ANALYZE SELECT * FROM tenk1 WHERE ten &lt; 7;
QUERY PLAN
-------------------------------------------------------------------&zwsp;--------------------------------------
Seq Scan on tenk1 (cost=0.00..483.00 rows=7000 width=244) (actual time=0.016..5.107 rows=7000 loops=1)
Filter: (ten &lt; 7)
Rows Removed by Filter: 3000
Planning time: 0.083 ms
Execution time: 5.905 ms
</screen>
These counts can be particularly valuable for filter conditions applied at
join nodes. The <quote>Rows Removed</quote> line only appears when at least
one scanned row, or potential join pair in the case of a join node,
is rejected by the filter condition.
</para>
<para>
A case similar to filter conditions occurs with <quote>lossy</quote>
index scans. For example, consider this search for polygons containing a
specific point:
<screen>
EXPLAIN ANALYZE SELECT * FROM polygon_tbl WHERE f1 @&gt; polygon '(0.5,2.0)';
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------------------------------
Seq Scan on polygon_tbl (cost=0.00..1.05 rows=1 width=32) (actual time=0.044..0.044 rows=0 loops=1)
Filter: (f1 @&gt; '((0.5,2))'::polygon)
Rows Removed by Filter: 4
Planning time: 0.040 ms
Execution time: 0.083 ms
</screen>
The planner thinks (quite correctly) that this sample table is too small
to bother with an index scan, so we have a plain sequential scan in which
all the rows got rejected by the filter condition. But if we force an
index scan to be used, we see:
<screen>
SET enable_seqscan TO off;
EXPLAIN ANALYZE SELECT * FROM polygon_tbl WHERE f1 @&gt; polygon '(0.5,2.0)';
QUERY PLAN
-------------------------------------------------------------------&zwsp;-------------------------------------------------------
Index Scan using gpolygonind on polygon_tbl (cost=0.13..8.15 rows=1 width=32) (actual time=0.062..0.062 rows=0 loops=1)
Index Cond: (f1 @&gt; '((0.5,2))'::polygon)
Rows Removed by Index Recheck: 1
Planning time: 0.034 ms
Execution time: 0.144 ms
</screen>
Here we can see that the index returned one candidate row, which was
then rejected by a recheck of the index condition. This happens because a
GiST index is <quote>lossy</quote> for polygon containment tests: it actually
returns the rows with polygons that overlap the target, and then we have
to do the exact containment test on those rows.
</para>
<para>
<command>EXPLAIN</command> has a <literal>BUFFERS</literal> option that can be used with
<literal>ANALYZE</literal> to get even more run time statistics:
<screen>
EXPLAIN (ANALYZE, BUFFERS) SELECT * FROM tenk1 WHERE unique1 &lt; 100 AND unique2 &gt; 9000;
QUERY PLAN
-------------------------------------------------------------------&zwsp;--------------------------------------------------------------
Bitmap Heap Scan on tenk1 (cost=25.08..60.21 rows=10 width=244) (actual time=0.323..0.342 rows=10 loops=1)
Recheck Cond: ((unique1 &lt; 100) AND (unique2 &gt; 9000))
Buffers: shared hit=15
-&gt; BitmapAnd (cost=25.08..25.08 rows=10 width=0) (actual time=0.309..0.309 rows=0 loops=1)
Buffers: shared hit=7
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..5.04 rows=101 width=0) (actual time=0.043..0.043 rows=100 loops=1)
Index Cond: (unique1 &lt; 100)
Buffers: shared hit=2
-&gt; Bitmap Index Scan on tenk1_unique2 (cost=0.00..19.78 rows=999 width=0) (actual time=0.227..0.227 rows=999 loops=1)
Index Cond: (unique2 &gt; 9000)
Buffers: shared hit=5
Planning time: 0.088 ms
Execution time: 0.423 ms
</screen>
The numbers provided by <literal>BUFFERS</literal> help to identify which parts
of the query are the most I/O-intensive.
</para>
<para>
Keep in mind that because <command>EXPLAIN ANALYZE</command> actually
runs the query, any side-effects will happen as usual, even though
whatever results the query might output are discarded in favor of
printing the <command>EXPLAIN</command> data. If you want to analyze a
data-modifying query without changing your tables, you can
roll the command back afterwards, for example:
<screen>
BEGIN;
EXPLAIN ANALYZE UPDATE tenk1 SET hundred = hundred + 1 WHERE unique1 &lt; 100;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-------------------------------------------------------------
Update on tenk1 (cost=5.08..230.08 rows=0 width=0) (actual time=3.791..3.792 rows=0 loops=1)
-&gt; Bitmap Heap Scan on tenk1 (cost=5.08..230.08 rows=102 width=10) (actual time=0.069..0.513 rows=100 loops=1)
Recheck Cond: (unique1 &lt; 100)
Heap Blocks: exact=90
-&gt; Bitmap Index Scan on tenk1_unique1 (cost=0.00..5.05 rows=102 width=0) (actual time=0.036..0.037 rows=300 loops=1)
Index Cond: (unique1 &lt; 100)
Planning Time: 0.113 ms
Execution Time: 3.850 ms
ROLLBACK;
</screen>
</para>
<para>
As seen in this example, when the query is an <command>INSERT</command>,
<command>UPDATE</command>, or <command>DELETE</command> command, the actual work of
applying the table changes is done by a top-level Insert, Update,
or Delete plan node. The plan nodes underneath this node perform
the work of locating the old rows and/or computing the new data.
So above, we see the same sort of bitmap table scan we've seen already,
and its output is fed to an Update node that stores the updated rows.
It's worth noting that although the data-modifying node can take a
considerable amount of run time (here, it's consuming the lion's share
of the time), the planner does not currently add anything to the cost
estimates to account for that work. That's because the work to be done is
the same for every correct query plan, so it doesn't affect planning
decisions.
</para>
<para>
When an <command>UPDATE</command> or <command>DELETE</command> command affects an
inheritance hierarchy, the output might look like this:
<screen>
EXPLAIN UPDATE parent SET f2 = f2 + 1 WHERE f1 = 101;
QUERY PLAN
-------------------------------------------------------------------&zwsp;-----------------------------------
Update on parent (cost=0.00..24.59 rows=0 width=0)
Update on parent parent_1
Update on child1 parent_2
Update on child2 parent_3
Update on child3 parent_4
-&gt; Result (cost=0.00..24.59 rows=4 width=14)
-&gt; Append (cost=0.00..24.54 rows=4 width=14)
-&gt; Seq Scan on parent parent_1 (cost=0.00..0.00 rows=1 width=14)
Filter: (f1 = 101)
-&gt; Index Scan using child1_pkey on child1 parent_2 (cost=0.15..8.17 rows=1 width=14)
Index Cond: (f1 = 101)
-&gt; Index Scan using child2_pkey on child2 parent_3 (cost=0.15..8.17 rows=1 width=14)
Index Cond: (f1 = 101)
-&gt; Index Scan using child3_pkey on child3 parent_4 (cost=0.15..8.17 rows=1 width=14)
Index Cond: (f1 = 101)
</screen>
In this example the Update node needs to consider three child tables as
well as the originally-mentioned parent table. So there are four input
scanning subplans, one per table. For clarity, the Update node is
annotated to show the specific target tables that will be updated, in the
same order as the corresponding subplans.
</para>
<para>
The <literal>Planning time</literal> shown by <command>EXPLAIN
ANALYZE</command> is the time it took to generate the query plan from the
parsed query and optimize it. It does not include parsing or rewriting.
</para>
<para>
The <literal>Execution time</literal> shown by <command>EXPLAIN
ANALYZE</command> includes executor start-up and shut-down time, as well
as the time to run any triggers that are fired, but it does not include
parsing, rewriting, or planning time.
Time spent executing <literal>BEFORE</literal> triggers, if any, is included in
the time for the related Insert, Update, or Delete node; but time
spent executing <literal>AFTER</literal> triggers is not counted there because
<literal>AFTER</literal> triggers are fired after completion of the whole plan.
The total time spent in each trigger
(either <literal>BEFORE</literal> or <literal>AFTER</literal>) is also shown separately.
Note that deferred constraint triggers will not be executed
until end of transaction and are thus not considered at all by
<command>EXPLAIN ANALYZE</command>.
</para>
</sect2>
<sect2 id="using-explain-caveats">
<title>Caveats</title>
<para>
There are two significant ways in which run times measured by
<command>EXPLAIN ANALYZE</command> can deviate from normal execution of
the same query. First, since no output rows are delivered to the client,
network transmission costs and I/O conversion costs are not included.
Second, the measurement overhead added by <command>EXPLAIN
ANALYZE</command> can be significant, especially on machines with slow
<function>gettimeofday()</function> operating-system calls. You can use the
<xref linkend="pgtesttiming"/> tool to measure the overhead of timing
on your system.
</para>
<para>
<command>EXPLAIN</command> results should not be extrapolated to situations
much different from the one you are actually testing; for example,
results on a toy-sized table cannot be assumed to apply to large tables.
The planner's cost estimates are not linear and so it might choose
a different plan for a larger or smaller table. An extreme example
is that on a table that only occupies one disk page, you'll nearly
always get a sequential scan plan whether indexes are available or not.
The planner realizes that it's going to take one disk page read to
process the table in any case, so there's no value in expending additional
page reads to look at an index. (We saw this happening in the
<literal>polygon_tbl</literal> example above.)
</para>
<para>
There are cases in which the actual and estimated values won't match up
well, but nothing is really wrong. One such case occurs when
plan node execution is stopped short by a <literal>LIMIT</literal> or similar
effect. For example, in the <literal>LIMIT</literal> query we used before,
<screen>
EXPLAIN ANALYZE SELECT * FROM tenk1 WHERE unique1 &lt; 100 AND unique2 &gt; 9000 LIMIT 2;
QUERY PLAN
-------------------------------------------------------------------&zwsp;------------------------------------------------------------
Limit (cost=0.29..14.71 rows=2 width=244) (actual time=0.177..0.249 rows=2 loops=1)
-&gt; Index Scan using tenk1_unique2 on tenk1 (cost=0.29..72.42 rows=10 width=244) (actual time=0.174..0.244 rows=2 loops=1)
Index Cond: (unique2 &gt; 9000)
Filter: (unique1 &lt; 100)
Rows Removed by Filter: 287
Planning time: 0.096 ms
Execution time: 0.336 ms
</screen>
the estimated cost and row count for the Index Scan node are shown as
though it were run to completion. But in reality the Limit node stopped
requesting rows after it got two, so the actual row count is only 2 and
the run time is less than the cost estimate would suggest. This is not
an estimation error, only a discrepancy in the way the estimates and true
values are displayed.
</para>
<para>
Merge joins also have measurement artifacts that can confuse the unwary.
A merge join will stop reading one input if it's exhausted the other input
and the next key value in the one input is greater than the last key value
of the other input; in such a case there can be no more matches and so no
need to scan the rest of the first input. This results in not reading all
of one child, with results like those mentioned for <literal>LIMIT</literal>.
Also, if the outer (first) child contains rows with duplicate key values,
the inner (second) child is backed up and rescanned for the portion of its
rows matching that key value. <command>EXPLAIN ANALYZE</command> counts these
repeated emissions of the same inner rows as if they were real additional
rows. When there are many outer duplicates, the reported actual row count
for the inner child plan node can be significantly larger than the number
of rows that are actually in the inner relation.
</para>
<para>
BitmapAnd and BitmapOr nodes always report their actual row counts as zero,
due to implementation limitations.
</para>
<para>
Normally, <command>EXPLAIN</command> will display every plan node
created by the planner. However, there are cases where the executor
can determine that certain nodes need not be executed because they
cannot produce any rows, based on parameter values that were not
available at planning time. (Currently this can only happen for child
nodes of an Append or MergeAppend node that is scanning a partitioned
table.) When this happens, those plan nodes are omitted from
the <command>EXPLAIN</command> output and a <literal>Subplans
Removed: <replaceable>N</replaceable></literal> annotation appears
instead.
</para>
</sect2>
</sect1>
<sect1 id="planner-stats">
<title>Statistics Used by the Planner</title>
<indexterm zone="planner-stats">
<primary>statistics</primary>
<secondary>of the planner</secondary>
</indexterm>
<sect2>
<title>Single-Column Statistics</title>
<para>
As we saw in the previous section, the query planner needs to estimate
the number of rows retrieved by a query in order to make good choices
of query plans. This section provides a quick look at the statistics
that the system uses for these estimates.
</para>
<para>
One component of the statistics is the total number of entries in
each table and index, as well as the number of disk blocks occupied
by each table and index. This information is kept in the table
<link linkend="catalog-pg-class"><structname>pg_class</structname></link>,
in the columns <structfield>reltuples</structfield> and
<structfield>relpages</structfield>. We can look at it with
queries similar to this one:
<screen>
SELECT relname, relkind, reltuples, relpages
FROM pg_class
WHERE relname LIKE 'tenk1%';
relname | relkind | reltuples | relpages
----------------------+---------+-----------+----------
tenk1 | r | 10000 | 358
tenk1_hundred | i | 10000 | 30
tenk1_thous_tenthous | i | 10000 | 30
tenk1_unique1 | i | 10000 | 30
tenk1_unique2 | i | 10000 | 30
(5 rows)
</screen>
Here we can see that <structname>tenk1</structname> contains 10000
rows, as do its indexes, but the indexes are (unsurprisingly) much
smaller than the table.
</para>
<para>
For efficiency reasons, <structfield>reltuples</structfield>
and <structfield>relpages</structfield> are not updated on-the-fly,
and so they usually contain somewhat out-of-date values.
They are updated by <command>VACUUM</command>, <command>ANALYZE</command>, and a
few DDL commands such as <command>CREATE INDEX</command>. A <command>VACUUM</command>
or <command>ANALYZE</command> operation that does not scan the entire table
(which is commonly the case) will incrementally update the
<structfield>reltuples</structfield> count on the basis of the part
of the table it did scan, resulting in an approximate value.
In any case, the planner
will scale the values it finds in <structname>pg_class</structname>
to match the current physical table size, thus obtaining a closer
approximation.
</para>
<indexterm>
<primary>pg_statistic</primary>
</indexterm>
<para>
Most queries retrieve only a fraction of the rows in a table, due
to <literal>WHERE</literal> clauses that restrict the rows to be
examined. The planner thus needs to make an estimate of the
<firstterm>selectivity</firstterm> of <literal>WHERE</literal> clauses, that is,
the fraction of rows that match each condition in the
<literal>WHERE</literal> clause. The information used for this task is
stored in the
<link linkend="catalog-pg-statistic"><structname>pg_statistic</structname></link>
system catalog. Entries in <structname>pg_statistic</structname>
are updated by the <command>ANALYZE</command> and <command>VACUUM
ANALYZE</command> commands, and are always approximate even when freshly
updated.
</para>
<indexterm>
<primary>pg_stats</primary>
</indexterm>
<para>
Rather than look at <structname>pg_statistic</structname> directly,
it's better to look at its view
<link linkend="view-pg-stats"><structname>pg_stats</structname></link>
when examining the statistics manually. <structname>pg_stats</structname>
is designed to be more easily readable. Furthermore,
<structname>pg_stats</structname> is readable by all, whereas
<structname>pg_statistic</structname> is only readable by a superuser.
(This prevents unprivileged users from learning something about
the contents of other people's tables from the statistics. The
<structname>pg_stats</structname> view is restricted to show only
rows about tables that the current user can read.)
For example, we might do:
<screen>
SELECT attname, inherited, n_distinct,
array_to_string(most_common_vals, E'\n') as most_common_vals
FROM pg_stats
WHERE tablename = 'road';
attname | inherited | n_distinct | most_common_vals
---------+-----------+------------+------------------------------------
name | f | -0.363388 | I- 580 Ramp+
| | | I- 880 Ramp+
| | | Sp Railroad +
| | | I- 580 +
| | | I- 680 Ramp
name | t | -0.284859 | I- 880 Ramp+
| | | I- 580 Ramp+
| | | I- 680 Ramp+
| | | I- 580 +
| | | State Hwy 13 Ramp
(2 rows)
</screen>
Note that two rows are displayed for the same column, one corresponding
to the complete inheritance hierarchy starting at the
<literal>road</literal> table (<literal>inherited</literal>=<literal>t</literal>),
and another one including only the <literal>road</literal> table itself
(<literal>inherited</literal>=<literal>f</literal>).
</para>
<para>
The amount of information stored in <structname>pg_statistic</structname>
by <command>ANALYZE</command>, in particular the maximum number of entries in the
<structfield>most_common_vals</structfield> and <structfield>histogram_bounds</structfield>
arrays for each column, can be set on a
column-by-column basis using the <command>ALTER TABLE SET STATISTICS</command>
command, or globally by setting the
<xref linkend="guc-default-statistics-target"/> configuration variable.
The default limit is presently 100 entries. Raising the limit
might allow more accurate planner estimates to be made, particularly for
columns with irregular data distributions, at the price of consuming
more space in <structname>pg_statistic</structname> and slightly more
time to compute the estimates. Conversely, a lower limit might be
sufficient for columns with simple data distributions.
</para>
<para>
Further details about the planner's use of statistics can be found in
<xref linkend="planner-stats-details"/>.
</para>
</sect2>
<sect2 id="planner-stats-extended">
<title>Extended Statistics</title>
<indexterm zone="planner-stats-extended">
<primary>statistics</primary>
<secondary>of the planner</secondary>
</indexterm>
<indexterm>
<primary>correlation</primary>
<secondary>in the query planner</secondary>
</indexterm>
<indexterm>
<primary>pg_statistic_ext</primary>
</indexterm>
<indexterm>
<primary>pg_statistic_ext_data</primary>
</indexterm>
<para>
It is common to see slow queries running bad execution plans because
multiple columns used in the query clauses are correlated.
The planner normally assumes that multiple conditions
are independent of each other,
an assumption that does not hold when column values are correlated.
Regular statistics, because of their per-individual-column nature,
cannot capture any knowledge about cross-column correlation.
However, <productname>PostgreSQL</productname> has the ability to compute
<firstterm>multivariate statistics</firstterm>, which can capture
such information.
</para>
<para>
Because the number of possible column combinations is very large,
it's impractical to compute multivariate statistics automatically.
Instead, <firstterm>extended statistics objects</firstterm>, more often
called just <firstterm>statistics objects</firstterm>, can be created to instruct
the server to obtain statistics across interesting sets of columns.
</para>
<para>
Statistics objects are created using the
<link linkend="sql-createstatistics"><command>CREATE STATISTICS</command></link> command.
Creation of such an object merely creates a catalog entry expressing
interest in the statistics. Actual data collection is performed
by <command>ANALYZE</command> (either a manual command, or background
auto-analyze). The collected values can be examined in the
<link linkend="catalog-pg-statistic-ext-data"><structname>pg_statistic_ext_data</structname></link>
catalog.
</para>
<para>
<command>ANALYZE</command> computes extended statistics based on the same
sample of table rows that it takes for computing regular single-column
statistics. Since the sample size is increased by increasing the
statistics target for the table or any of its columns (as described in
the previous section), a larger statistics target will normally result in
more accurate extended statistics, as well as more time spent calculating
them.
</para>
<para>
The following subsections describe the kinds of extended statistics
that are currently supported.
</para>
<sect3>
<title>Functional Dependencies</title>
<para>
The simplest kind of extended statistics tracks <firstterm>functional
dependencies</firstterm>, a concept used in definitions of database normal forms.
We say that column <structfield>b</structfield> is functionally dependent on
column <structfield>a</structfield> if knowledge of the value of
<structfield>a</structfield> is sufficient to determine the value
of <structfield>b</structfield>, that is there are no two rows having the same value
of <structfield>a</structfield> but different values of <structfield>b</structfield>.
In a fully normalized database, functional dependencies should exist
only on primary keys and superkeys. However, in practice many data sets
are not fully normalized for various reasons; intentional
denormalization for performance reasons is a common example.
Even in a fully normalized database, there may be partial correlation
between some columns, which can be expressed as partial functional
dependency.
</para>
<para>
The existence of functional dependencies directly affects the accuracy
of estimates in certain queries. If a query contains conditions on
both the independent and the dependent column(s), the
conditions on the dependent columns do not further reduce the result
size; but without knowledge of the functional dependency, the query
planner will assume that the conditions are independent, resulting
in underestimating the result size.
</para>
<para>
To inform the planner about functional dependencies, <command>ANALYZE</command>
can collect measurements of cross-column dependency. Assessing the
degree of dependency between all sets of columns would be prohibitively
expensive, so data collection is limited to those groups of columns
appearing together in a statistics object defined with
the <literal>dependencies</literal> option. It is advisable to create
<literal>dependencies</literal> statistics only for column groups that are
strongly correlated, to avoid unnecessary overhead in both
<command>ANALYZE</command> and later query planning.
</para>
<para>
Here is an example of collecting functional-dependency statistics:
<programlisting>
CREATE STATISTICS stts (dependencies) ON city, zip FROM zipcodes;
ANALYZE zipcodes;
SELECT stxname, stxkeys, stxddependencies
FROM pg_statistic_ext join pg_statistic_ext_data on (oid = stxoid)
WHERE stxname = 'stts';
stxname | stxkeys | stxddependencies
---------+---------+------------------------------------------
stts | 1 5 | {"1 => 5": 1.000000, "5 => 1": 0.423130}
(1 row)
</programlisting>
Here it can be seen that column 1 (zip code) fully determines column
5 (city) so the coefficient is 1.0, while city only determines zip code
about 42% of the time, meaning that there are many cities (58%) that are
represented by more than a single ZIP code.
</para>
<para>
When computing the selectivity for a query involving functionally
dependent columns, the planner adjusts the per-condition selectivity
estimates using the dependency coefficients so as not to produce
an underestimate.
</para>
<sect4>
<title>Limitations of Functional Dependencies</title>
<para>
Functional dependencies are currently only applied when considering
simple equality conditions that compare columns to constant values,
and <literal>IN</literal> clauses with constant values.
They are not used to improve estimates for equality conditions
comparing two columns or comparing a column to an expression, nor for
range clauses, <literal>LIKE</literal> or any other type of condition.
</para>
<para>
When estimating with functional dependencies, the planner assumes that
conditions on the involved columns are compatible and hence redundant.
If they are incompatible, the correct estimate would be zero rows, but
that possibility is not considered. For example, given a query like
<programlisting>
SELECT * FROM zipcodes WHERE city = 'San Francisco' AND zip = '94105';
</programlisting>
the planner will disregard the <structfield>city</structfield> clause as not
changing the selectivity, which is correct. However, it will make
the same assumption about
<programlisting>
SELECT * FROM zipcodes WHERE city = 'San Francisco' AND zip = '90210';
</programlisting>
even though there will really be zero rows satisfying this query.
Functional dependency statistics do not provide enough information
to conclude that, however.
</para>
<para>
In many practical situations, this assumption is usually satisfied;
for example, there might be a GUI in the application that only allows
selecting compatible city and ZIP code values to use in a query.
But if that's not the case, functional dependencies may not be a viable
option.
</para>
</sect4>
</sect3>
<sect3>
<title>Multivariate N-Distinct Counts</title>
<para>
Single-column statistics store the number of distinct values in each
column. Estimates of the number of distinct values when combining more
than one column (for example, for <literal>GROUP BY a, b</literal>) are
frequently wrong when the planner only has single-column statistical
data, causing it to select bad plans.
</para>
<para>
To improve such estimates, <command>ANALYZE</command> can collect n-distinct
statistics for groups of columns. As before, it's impractical to do
this for every possible column grouping, so data is collected only for
those groups of columns appearing together in a statistics object
defined with the <literal>ndistinct</literal> option. Data will be collected
for each possible combination of two or more columns from the set of
listed columns.
</para>
<para>
Continuing the previous example, the n-distinct counts in a
table of ZIP codes might look like the following:
<programlisting>
CREATE STATISTICS stts2 (ndistinct) ON city, state, zip FROM zipcodes;
ANALYZE zipcodes;
SELECT stxkeys AS k, stxdndistinct AS nd
FROM pg_statistic_ext join pg_statistic_ext_data on (oid = stxoid)
WHERE stxname = 'stts2';
-[ RECORD 1 ]------------------------------------------------------&zwsp;--
k | 1 2 5
nd | {"1, 2": 33178, "1, 5": 33178, "2, 5": 27435, "1, 2, 5": 33178}
(1 row)
</programlisting>
This indicates that there are three combinations of columns that
have 33178 distinct values: ZIP code and state; ZIP code and city;
and ZIP code, city and state (the fact that they are all equal is
expected given that ZIP code alone is unique in this table). On the
other hand, the combination of city and state has only 27435 distinct
values.
</para>
<para>
It's advisable to create <literal>ndistinct</literal> statistics objects only
on combinations of columns that are actually used for grouping, and
for which misestimation of the number of groups is resulting in bad
plans. Otherwise, the <command>ANALYZE</command> cycles are just wasted.
</para>
</sect3>
<sect3>
<title>Multivariate MCV Lists</title>
<para>
Another type of statistic stored for each column are most-common value
lists. This allows very accurate estimates for individual columns, but
may result in significant misestimates for queries with conditions on
multiple columns.
</para>
<para>
To improve such estimates, <command>ANALYZE</command> can collect MCV
lists on combinations of columns. Similarly to functional dependencies
and n-distinct coefficients, it's impractical to do this for every
possible column grouping. Even more so in this case, as the MCV list
(unlike functional dependencies and n-distinct coefficients) does store
the common column values. So data is collected only for those groups
of columns appearing together in a statistics object defined with the
<literal>mcv</literal> option.
</para>
<para>
Continuing the previous example, the MCV list for a table of ZIP codes
might look like the following (unlike for simpler types of statistics,
a function is required for inspection of MCV contents):
<programlisting>
CREATE STATISTICS stts3 (mcv) ON city, state FROM zipcodes;
ANALYZE zipcodes;
SELECT m.* FROM pg_statistic_ext join pg_statistic_ext_data on (oid = stxoid),
pg_mcv_list_items(stxdmcv) m WHERE stxname = 'stts3';
index | values | nulls | frequency | base_frequency
-------+------------------------+-------+-----------+----------------
0 | {Washington, DC} | {f,f} | 0.003467 | 2.7e-05
1 | {Apo, AE} | {f,f} | 0.003067 | 1.9e-05
2 | {Houston, TX} | {f,f} | 0.002167 | 0.000133
3 | {El Paso, TX} | {f,f} | 0.002 | 0.000113
4 | {New York, NY} | {f,f} | 0.001967 | 0.000114
5 | {Atlanta, GA} | {f,f} | 0.001633 | 3.3e-05
6 | {Sacramento, CA} | {f,f} | 0.001433 | 7.8e-05
7 | {Miami, FL} | {f,f} | 0.0014 | 6e-05
8 | {Dallas, TX} | {f,f} | 0.001367 | 8.8e-05
9 | {Chicago, IL} | {f,f} | 0.001333 | 5.1e-05
...
(99 rows)
</programlisting>
This indicates that the most common combination of city and state is
Washington in DC, with actual frequency (in the sample) about 0.35%.
The base frequency of the combination (as computed from the simple
per-column frequencies) is only 0.0027%, resulting in two orders of
magnitude under-estimates.
</para>
<para>
It's advisable to create <acronym>MCV</acronym> statistics objects only
on combinations of columns that are actually used in conditions together,
and for which misestimation of the number of groups is resulting in bad
plans. Otherwise, the <command>ANALYZE</command> and planning cycles
are just wasted.
</para>
</sect3>
</sect2>
</sect1>
<sect1 id="explicit-joins">
<title>Controlling the Planner with Explicit <literal>JOIN</literal> Clauses</title>
<indexterm zone="explicit-joins">
<primary>join</primary>
<secondary>controlling the order</secondary>
</indexterm>
<para>
It is possible
to control the query planner to some extent by using the explicit <literal>JOIN</literal>
syntax. To see why this matters, we first need some background.
</para>
<para>
In a simple join query, such as:
<programlisting>
SELECT * FROM a, b, c WHERE a.id = b.id AND b.ref = c.id;
</programlisting>
the planner is free to join the given tables in any order. For
example, it could generate a query plan that joins A to B, using
the <literal>WHERE</literal> condition <literal>a.id = b.id</literal>, and then
joins C to this joined table, using the other <literal>WHERE</literal>
condition. Or it could join B to C and then join A to that result.
Or it could join A to C and then join them with B &mdash; but that
would be inefficient, since the full Cartesian product of A and C
would have to be formed, there being no applicable condition in the
<literal>WHERE</literal> clause to allow optimization of the join. (All
joins in the <productname>PostgreSQL</productname> executor happen
between two input tables, so it's necessary to build up the result
in one or another of these fashions.) The important point is that
these different join possibilities give semantically equivalent
results but might have hugely different execution costs. Therefore,
the planner will explore all of them to try to find the most
efficient query plan.
</para>
<para>
When a query only involves two or three tables, there aren't many join
orders to worry about. But the number of possible join orders grows
exponentially as the number of tables expands. Beyond ten or so input
tables it's no longer practical to do an exhaustive search of all the
possibilities, and even for six or seven tables planning might take an
annoyingly long time. When there are too many input tables, the
<productname>PostgreSQL</productname> planner will switch from exhaustive
search to a <firstterm>genetic</firstterm> probabilistic search
through a limited number of possibilities. (The switch-over threshold is
set by the <xref linkend="guc-geqo-threshold"/> run-time
parameter.)
The genetic search takes less time, but it won't
necessarily find the best possible plan.
</para>
<para>
When the query involves outer joins, the planner has less freedom
than it does for plain (inner) joins. For example, consider:
<programlisting>
SELECT * FROM a LEFT JOIN (b JOIN c ON (b.ref = c.id)) ON (a.id = b.id);
</programlisting>
Although this query's restrictions are superficially similar to the
previous example, the semantics are different because a row must be
emitted for each row of A that has no matching row in the join of B and C.
Therefore the planner has no choice of join order here: it must join
B to C and then join A to that result. Accordingly, this query takes
less time to plan than the previous query. In other cases, the planner
might be able to determine that more than one join order is safe.
For example, given:
<programlisting>
SELECT * FROM a LEFT JOIN b ON (a.bid = b.id) LEFT JOIN c ON (a.cid = c.id);
</programlisting>
it is valid to join A to either B or C first. Currently, only
<literal>FULL JOIN</literal> completely constrains the join order. Most
practical cases involving <literal>LEFT JOIN</literal> or <literal>RIGHT JOIN</literal>
can be rearranged to some extent.
</para>
<para>
Explicit inner join syntax (<literal>INNER JOIN</literal>, <literal>CROSS
JOIN</literal>, or unadorned <literal>JOIN</literal>) is semantically the same as
listing the input relations in <literal>FROM</literal>, so it does not
constrain the join order.
</para>
<para>
Even though most kinds of <literal>JOIN</literal> don't completely constrain
the join order, it is possible to instruct the
<productname>PostgreSQL</productname> query planner to treat all
<literal>JOIN</literal> clauses as constraining the join order anyway.
For example, these three queries are logically equivalent:
<programlisting>
SELECT * FROM a, b, c WHERE a.id = b.id AND b.ref = c.id;
SELECT * FROM a CROSS JOIN b CROSS JOIN c WHERE a.id = b.id AND b.ref = c.id;
SELECT * FROM a JOIN (b JOIN c ON (b.ref = c.id)) ON (a.id = b.id);
</programlisting>
But if we tell the planner to honor the <literal>JOIN</literal> order,
the second and third take less time to plan than the first. This effect
is not worth worrying about for only three tables, but it can be a
lifesaver with many tables.
</para>
<para>
To force the planner to follow the join order laid out by explicit
<literal>JOIN</literal>s,
set the <xref linkend="guc-join-collapse-limit"/> run-time parameter to 1.
(Other possible values are discussed below.)
</para>
<para>
You do not need to constrain the join order completely in order to
cut search time, because it's OK to use <literal>JOIN</literal> operators
within items of a plain <literal>FROM</literal> list. For example, consider:
<programlisting>
SELECT * FROM a CROSS JOIN b, c, d, e WHERE ...;
</programlisting>
With <varname>join_collapse_limit</varname> = 1, this
forces the planner to join A to B before joining them to other tables,
but doesn't constrain its choices otherwise. In this example, the
number of possible join orders is reduced by a factor of 5.
</para>
<para>
Constraining the planner's search in this way is a useful technique
both for reducing planning time and for directing the planner to a
good query plan. If the planner chooses a bad join order by default,
you can force it to choose a better order via <literal>JOIN</literal> syntax
&mdash; assuming that you know of a better order, that is. Experimentation
is recommended.
</para>
<para>
A closely related issue that affects planning time is collapsing of
subqueries into their parent query. For example, consider:
<programlisting>
SELECT *
FROM x, y,
(SELECT * FROM a, b, c WHERE something) AS ss
WHERE somethingelse;
</programlisting>
This situation might arise from use of a view that contains a join;
the view's <literal>SELECT</literal> rule will be inserted in place of the view
reference, yielding a query much like the above. Normally, the planner
will try to collapse the subquery into the parent, yielding:
<programlisting>
SELECT * FROM x, y, a, b, c WHERE something AND somethingelse;
</programlisting>
This usually results in a better plan than planning the subquery
separately. (For example, the outer <literal>WHERE</literal> conditions might be such that
joining X to A first eliminates many rows of A, thus avoiding the need to
form the full logical output of the subquery.) But at the same time,
we have increased the planning time; here, we have a five-way join
problem replacing two separate three-way join problems. Because of the
exponential growth of the number of possibilities, this makes a big
difference. The planner tries to avoid getting stuck in huge join search
problems by not collapsing a subquery if more than <varname>from_collapse_limit</varname>
<literal>FROM</literal> items would result in the parent
query. You can trade off planning time against quality of plan by
adjusting this run-time parameter up or down.
</para>
<para>
<xref linkend="guc-from-collapse-limit"/> and <xref
linkend="guc-join-collapse-limit"/>
are similarly named because they do almost the same thing: one controls
when the planner will <quote>flatten out</quote> subqueries, and the
other controls when it will flatten out explicit joins. Typically
you would either set <varname>join_collapse_limit</varname> equal to
<varname>from_collapse_limit</varname> (so that explicit joins and subqueries
act similarly) or set <varname>join_collapse_limit</varname> to 1 (if you want
to control join order with explicit joins). But you might set them
differently if you are trying to fine-tune the trade-off between planning
time and run time.
</para>
</sect1>
<sect1 id="populate">
<title>Populating a Database</title>
<para>
One might need to insert a large amount of data when first populating
a database. This section contains some suggestions on how to make
this process as efficient as possible.
</para>
<sect2 id="disable-autocommit">
<title>Disable Autocommit</title>
<indexterm>
<primary>autocommit</primary>
<secondary>bulk-loading data</secondary>
</indexterm>
<para>
When using multiple <command>INSERT</command>s, turn off autocommit and just do
one commit at the end. (In plain
SQL, this means issuing <command>BEGIN</command> at the start and
<command>COMMIT</command> at the end. Some client libraries might
do this behind your back, in which case you need to make sure the
library does it when you want it done.) If you allow each
insertion to be committed separately,
<productname>PostgreSQL</productname> is doing a lot of work for
each row that is added. An additional benefit of doing all
insertions in one transaction is that if the insertion of one row
were to fail then the insertion of all rows inserted up to that
point would be rolled back, so you won't be stuck with partially
loaded data.
</para>
</sect2>
<sect2 id="populate-copy-from">
<title>Use <command>COPY</command></title>
<para>
Use <link linkend="sql-copy"><command>COPY</command></link> to load
all the rows in one command, instead of using a series of
<command>INSERT</command> commands. The <command>COPY</command>
command is optimized for loading large numbers of rows; it is less
flexible than <command>INSERT</command>, but incurs significantly
less overhead for large data loads. Since <command>COPY</command>
is a single command, there is no need to disable autocommit if you
use this method to populate a table.
</para>
<para>
If you cannot use <command>COPY</command>, it might help to use <link
linkend="sql-prepare"><command>PREPARE</command></link> to create a
prepared <command>INSERT</command> statement, and then use
<command>EXECUTE</command> as many times as required. This avoids
some of the overhead of repeatedly parsing and planning
<command>INSERT</command>. Different interfaces provide this facility
in different ways; look for <quote>prepared statements</quote> in the interface
documentation.
</para>
<para>
Note that loading a large number of rows using
<command>COPY</command> is almost always faster than using
<command>INSERT</command>, even if <command>PREPARE</command> is used and
multiple insertions are batched into a single transaction.
</para>
<para>
<command>COPY</command> is fastest when used within the same
transaction as an earlier <command>CREATE TABLE</command> or
<command>TRUNCATE</command> command. In such cases no WAL
needs to be written, because in case of an error, the files
containing the newly loaded data will be removed anyway.
However, this consideration only applies when
<xref linkend="guc-wal-level"/> is <literal>minimal</literal>
as all commands must write WAL otherwise.
</para>
</sect2>
<sect2 id="populate-rm-indexes">
<title>Remove Indexes</title>
<para>
If you are loading a freshly created table, the fastest method is to
create the table, bulk load the table's data using
<command>COPY</command>, then create any indexes needed for the
table. Creating an index on pre-existing data is quicker than
updating it incrementally as each row is loaded.
</para>
<para>
If you are adding large amounts of data to an existing table,
it might be a win to drop the indexes,
load the table, and then recreate the indexes. Of course, the
database performance for other users might suffer
during the time the indexes are missing. One should also think
twice before dropping a unique index, since the error checking
afforded by the unique constraint will be lost while the index is
missing.
</para>
</sect2>
<sect2 id="populate-rm-fkeys">
<title>Remove Foreign Key Constraints</title>
<para>
Just as with indexes, a foreign key constraint can be checked
<quote>in bulk</quote> more efficiently than row-by-row. So it might be
useful to drop foreign key constraints, load data, and re-create
the constraints. Again, there is a trade-off between data load
speed and loss of error checking while the constraint is missing.
</para>
<para>
What's more, when you load data into a table with existing foreign key
constraints, each new row requires an entry in the server's list of
pending trigger events (since it is the firing of a trigger that checks
the row's foreign key constraint). Loading many millions of rows can
cause the trigger event queue to overflow available memory, leading to
intolerable swapping or even outright failure of the command. Therefore
it may be <emphasis>necessary</emphasis>, not just desirable, to drop and re-apply
foreign keys when loading large amounts of data. If temporarily removing
the constraint isn't acceptable, the only other recourse may be to split
up the load operation into smaller transactions.
</para>
</sect2>
<sect2 id="populate-work-mem">
<title>Increase <varname>maintenance_work_mem</varname></title>
<para>
Temporarily increasing the <xref linkend="guc-maintenance-work-mem"/>
configuration variable when loading large amounts of data can
lead to improved performance. This will help to speed up <command>CREATE
INDEX</command> commands and <command>ALTER TABLE ADD FOREIGN KEY</command> commands.
It won't do much for <command>COPY</command> itself, so this advice is
only useful when you are using one or both of the above techniques.
</para>
</sect2>
<sect2 id="populate-max-wal-size">
<title>Increase <varname>max_wal_size</varname></title>
<para>
Temporarily increasing the <xref linkend="guc-max-wal-size"/>
configuration variable can also
make large data loads faster. This is because loading a large
amount of data into <productname>PostgreSQL</productname> will
cause checkpoints to occur more often than the normal checkpoint
frequency (specified by the <varname>checkpoint_timeout</varname>
configuration variable). Whenever a checkpoint occurs, all dirty
pages must be flushed to disk. By increasing
<varname>max_wal_size</varname> temporarily during bulk
data loads, the number of checkpoints that are required can be
reduced.
</para>
</sect2>
<sect2 id="populate-pitr">
<title>Disable WAL Archival and Streaming Replication</title>
<para>
When loading large amounts of data into an installation that uses
WAL archiving or streaming replication, it might be faster to take a
new base backup after the load has completed than to process a large
amount of incremental WAL data. To prevent incremental WAL logging
while loading, disable archiving and streaming replication, by setting
<xref linkend="guc-wal-level"/> to <literal>minimal</literal>,
<xref linkend="guc-archive-mode"/> to <literal>off</literal>, and
<xref linkend="guc-max-wal-senders"/> to zero.
But note that changing these settings requires a server restart,
and makes any base backups taken before unavailable for archive
recovery and standby server, which may lead to data loss.
</para>
<para>
Aside from avoiding the time for the archiver or WAL sender to process the
WAL data, doing this will actually make certain commands faster, because
they do not to write WAL at all if <varname>wal_level</varname>
is <literal>minimal</literal> and the current subtransaction (or top-level
transaction) created or truncated the table or index they change. (They
can guarantee crash safety more cheaply by doing
an <function>fsync</function> at the end than by writing WAL.)
</para>
</sect2>
<sect2 id="populate-analyze">
<title>Run <command>ANALYZE</command> Afterwards</title>
<para>
Whenever you have significantly altered the distribution of data
within a table, running <link linkend="sql-analyze"><command>ANALYZE</command></link> is strongly recommended. This
includes bulk loading large amounts of data into the table. Running
<command>ANALYZE</command> (or <command>VACUUM ANALYZE</command>)
ensures that the planner has up-to-date statistics about the
table. With no statistics or obsolete statistics, the planner might
make poor decisions during query planning, leading to poor
performance on any tables with inaccurate or nonexistent
statistics. Note that if the autovacuum daemon is enabled, it might
run <command>ANALYZE</command> automatically; see
<xref linkend="vacuum-for-statistics"/>
and <xref linkend="autovacuum"/> for more information.
</para>
</sect2>
<sect2 id="populate-pg-dump">
<title>Some Notes about <application>pg_dump</application></title>
<para>
Dump scripts generated by <application>pg_dump</application> automatically apply
several, but not all, of the above guidelines. To restore a
<application>pg_dump</application> dump as quickly as possible, you need to
do a few extra things manually. (Note that these points apply while
<emphasis>restoring</emphasis> a dump, not while <emphasis>creating</emphasis> it.
The same points apply whether loading a text dump with
<application>psql</application> or using <application>pg_restore</application> to load
from a <application>pg_dump</application> archive file.)
</para>
<para>
By default, <application>pg_dump</application> uses <command>COPY</command>, and when
it is generating a complete schema-and-data dump, it is careful to
load data before creating indexes and foreign keys. So in this case
several guidelines are handled automatically. What is left
for you to do is to:
<itemizedlist>
<listitem>
<para>
Set appropriate (i.e., larger than normal) values for
<varname>maintenance_work_mem</varname> and
<varname>max_wal_size</varname>.
</para>
</listitem>
<listitem>
<para>
If using WAL archiving or streaming replication, consider disabling
them during the restore. To do that, set <varname>archive_mode</varname>
to <literal>off</literal>,
<varname>wal_level</varname> to <literal>minimal</literal>, and
<varname>max_wal_senders</varname> to zero before loading the dump.
Afterwards, set them back to the right values and take a fresh
base backup.
</para>
</listitem>
<listitem>
<para>
Experiment with the parallel dump and restore modes of both
<application>pg_dump</application> and <application>pg_restore</application> and find the
optimal number of concurrent jobs to use. Dumping and restoring in
parallel by means of the <option>-j</option> option should give you a
significantly higher performance over the serial mode.
</para>
</listitem>
<listitem>
<para>
Consider whether the whole dump should be restored as a single
transaction. To do that, pass the <option>-1</option> or
<option>--single-transaction</option> command-line option to
<application>psql</application> or <application>pg_restore</application>. When using this
mode, even the smallest of errors will rollback the entire restore,
possibly discarding many hours of processing. Depending on how
interrelated the data is, that might seem preferable to manual cleanup,
or not. <command>COPY</command> commands will run fastest if you use a single
transaction and have WAL archiving turned off.
</para>
</listitem>
<listitem>
<para>
If multiple CPUs are available in the database server, consider using
<application>pg_restore</application>'s <option>--jobs</option> option. This
allows concurrent data loading and index creation.
</para>
</listitem>
<listitem>
<para>
Run <command>ANALYZE</command> afterwards.
</para>
</listitem>
</itemizedlist>
</para>
<para>
A data-only dump will still use <command>COPY</command>, but it does not
drop or recreate indexes, and it does not normally touch foreign
keys.
<footnote>
<para>
You can get the effect of disabling foreign keys by using
the <option>--disable-triggers</option> option &mdash; but realize that
that eliminates, rather than just postpones, foreign key
validation, and so it is possible to insert bad data if you use it.
</para>
</footnote>
So when loading a data-only dump, it is up to you to drop and recreate
indexes and foreign keys if you wish to use those techniques.
It's still useful to increase <varname>max_wal_size</varname>
while loading the data, but don't bother increasing
<varname>maintenance_work_mem</varname>; rather, you'd do that while
manually recreating indexes and foreign keys afterwards.
And don't forget to <command>ANALYZE</command> when you're done; see
<xref linkend="vacuum-for-statistics"/>
and <xref linkend="autovacuum"/> for more information.
</para>
</sect2>
</sect1>
<sect1 id="non-durability">
<title>Non-Durable Settings</title>
<indexterm zone="non-durability">
<primary>non-durable</primary>
</indexterm>
<para>
Durability is a database feature that guarantees the recording of
committed transactions even if the server crashes or loses
power. However, durability adds significant database overhead,
so if your site does not require such a guarantee,
<productname>PostgreSQL</productname> can be configured to run
much faster. The following are configuration changes you can make
to improve performance in such cases. Except as noted below, durability
is still guaranteed in case of a crash of the database software;
only an abrupt operating system crash creates a risk of data loss
or corruption when these settings are used.
<itemizedlist>
<listitem>
<para>
Place the database cluster's data directory in a memory-backed
file system (i.e., <acronym>RAM</acronym> disk). This eliminates all
database disk I/O, but limits data storage to the amount of
available memory (and perhaps swap).
</para>
</listitem>
<listitem>
<para>
Turn off <xref linkend="guc-fsync"/>; there is no need to flush
data to disk.
</para>
</listitem>
<listitem>
<para>
Turn off <xref linkend="guc-synchronous-commit"/>; there might be no
need to force <acronym>WAL</acronym> writes to disk on every
commit. This setting does risk transaction loss (though not data
corruption) in case of a crash of the <emphasis>database</emphasis>.
</para>
</listitem>
<listitem>
<para>
Turn off <xref linkend="guc-full-page-writes"/>; there is no need
to guard against partial page writes.
</para>
</listitem>
<listitem>
<para>
Increase <xref linkend="guc-max-wal-size"/> and <xref
linkend="guc-checkpoint-timeout"/>; this reduces the frequency
of checkpoints, but increases the storage requirements of
<filename>/pg_wal</filename>.
</para>
</listitem>
<listitem>
<para>
Create <link linkend="sql-createtable-unlogged">unlogged
tables</link> to avoid <acronym>WAL</acronym> writes, though it
makes the tables non-crash-safe.
</para>
</listitem>
</itemizedlist>
</para>
</sect1>
</chapter>