path: root/doc/src/sgml/ref/pgbench.sgml

<!-- doc/src/sgml/ref/pgbench.sgml -->

<refentry id="pgbench">
 <indexterm zone="pgbench">
  <primary>pgbench</primary>
 </indexterm>

 <refmeta>
  <refentrytitle><application>pgbench</application></refentrytitle>
  <manvolnum>1</manvolnum>
  <refmiscinfo>Application</refmiscinfo>
 </refmeta>

 <refnamediv>
  <refname>pgbench</refname>
  <refpurpose>run a benchmark test on <productname>PostgreSQL</productname></refpurpose>
 </refnamediv>

 <refsynopsisdiv>
  <cmdsynopsis>
   <command>pgbench</command>
   <arg choice="plain"><option>-i</option></arg>
   <arg rep="repeat"><replaceable>option</replaceable></arg>
   <arg choice="opt"><replaceable>dbname</replaceable></arg>
  </cmdsynopsis>
  <cmdsynopsis>
   <command>pgbench</command>
   <arg rep="repeat"><replaceable>option</replaceable></arg>
   <arg choice="opt"><replaceable>dbname</replaceable></arg>
  </cmdsynopsis>
 </refsynopsisdiv>

 <refsect1>
  <title>Description</title>
 <para>
  <application>pgbench</application> is a simple program for running benchmark
  tests on <productname>PostgreSQL</>.  It runs the same sequence of SQL
  commands over and over, possibly in multiple concurrent database sessions,
  and then calculates the average transaction rate (transactions per second).
  By default, <application>pgbench</application> tests a scenario that is
  loosely based on TPC-B, involving five <command>SELECT</>,
  <command>UPDATE</>, and <command>INSERT</> commands per transaction.
  However, it is easy to test other cases by writing your own transaction
  script files.
 </para>

 <para>
  Typical output from <application>pgbench</application> looks like:

<screen>
transaction type: &lt;builtin: TPC-B (sort of)&gt;
scaling factor: 10
query mode: simple
number of clients: 10
number of threads: 1
number of transactions per client: 1000
number of transactions actually processed: 10000/10000
tps = 85.184871 (including connections establishing)
tps = 85.296346 (excluding connections establishing)
</screen>

  The first six lines report some of the most important parameter
  settings.  The next line reports the number of transactions completed
  and intended (the latter being just the product of number of clients
  and number of transactions per client); these will be equal unless the run
  failed before completion.  (In <option>-T</> mode, only the actual
  number of transactions is printed.)
  The last two lines report the number of transactions per second,
  figured with and without counting the time to start database sessions.
 </para>

  <para>
   The default TPC-B-like transaction test requires specific tables to be
   set up beforehand.  <application>pgbench</> should be invoked with
   the <option>-i</> (initialize) option to create and populate these
   tables.  (When you are testing a custom script, you don't need this
   step, but will instead need to do whatever setup your test needs.)
   Initialization looks like:

<programlisting>
pgbench -i <optional> <replaceable>other-options</> </optional> <replaceable>dbname</>
</programlisting>

   where <replaceable>dbname</> is the name of the already-created
   database to test in.  (You may also need <option>-h</>,
   <option>-p</>, and/or <option>-U</> options to specify how to
   connect to the database server.)
  </para>

  <caution>
   <para>
    <literal>pgbench -i</> creates four tables <structname>pgbench_accounts</>,
    <structname>pgbench_branches</>, <structname>pgbench_history</>, and
    <structname>pgbench_tellers</>,
    destroying any existing tables of these names.
    Be very careful to use another database if you have tables having these
    names!
   </para>
  </caution>

  <para>
   At the default <quote>scale factor</> of 1, the tables initially
   contain this many rows:
<screen>
table                   # of rows
---------------------------------
pgbench_branches        1
pgbench_tellers         10
pgbench_accounts        100000
pgbench_history         0
</screen>
   You can (and, for most purposes, probably should) increase the number
   of rows by using the <option>-s</> (scale factor) option.  The
   <option>-F</> (fillfactor) option might also be used at this point.
  </para>

  <para>
   Once you have done the necessary setup, you can run your benchmark
   with a command that doesn't include <option>-i</>, that is

<programlisting>
pgbench <optional> <replaceable>options</> </optional> <replaceable>dbname</>
</programlisting>

   In nearly all cases, you'll need some options to make a useful test.
   The most important options are <option>-c</> (number of clients),
   <option>-t</> (number of transactions), <option>-T</> (time limit),
   and <option>-f</> (specify a custom script file).
   See below for a full list.
  </para>
 </refsect1>

 <refsect1>
  <title>Options</title>

  <para>
   The following is divided into three subsections: Different options are used
   during database initialization and while running benchmarks, some options
   are useful in both cases.
  </para>

 <refsect2 id="pgbench-init-options">
  <title>Initialization Options</title>

   <para>
    <application>pgbench</application> accepts the following command-line
    initialization arguments:

    <variablelist>

     <varlistentry>
      <term><option>-i</option></term>
      <term><option>--initialize</option></term>
      <listitem>
       <para>
        Required to invoke initialization mode.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-F</option> <replaceable>fillfactor</></term>
      <term><option>--fillfactor=</option><replaceable>fillfactor</></term>
      <listitem>
       <para>
        Create the <structname>pgbench_accounts</>,
        <structname>pgbench_tellers</> and
        <structname>pgbench_branches</> tables with the given fillfactor.
        Default is 100.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-n</option></term>
      <term><option>--no-vacuum</option></term>
      <listitem>
       <para>
        Perform no vacuuming after initialization.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-q</option></term>
      <term><option>--quiet</option></term>
      <listitem>
       <para>
        Switch logging to quiet mode, producing only one progress message per 5
        seconds. The default logging prints one message each 100000 rows, which
        often outputs many lines per second (especially on good hardware).
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-s</option> <replaceable>scale_factor</></term>
      <term><option>--scale=</option><replaceable>scale_factor</></term>
      <listitem>
       <para>
        Multiply the number of rows generated by the scale factor.
        For example, <literal>-s 100</> will create 10,000,000 rows
        in the <structname>pgbench_accounts</> table. Default is 1.
        When the scale is 20,000 or larger, the columns used to
        hold account identifiers (<structfield>aid</structfield> columns)
        will switch to using larger integers (<type>bigint</type>),
        in order to be big enough to hold the range of account
        identifiers.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--foreign-keys</option></term>
      <listitem>
       <para>
        Create foreign key constraints between the standard tables.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--index-tablespace=<replaceable>index_tablespace</replaceable></option></term>
      <listitem>
       <para>
        Create indexes in the specified tablespace, rather than the default
        tablespace.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--tablespace=<replaceable>tablespace</replaceable></option></term>
      <listitem>
       <para>
        Create tables in the specified tablespace, rather than the default
        tablespace.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--unlogged-tables</option></term>
      <listitem>
       <para>
        Create all tables as unlogged tables, rather than permanent tables.
       </para>
      </listitem>
     </varlistentry>

    </variablelist>
   </para>

 </refsect2>

 <refsect2 id="pgbench-run-options">
  <title>Benchmarking Options</title>

   <para>
    <application>pgbench</application> accepts the following command-line
    benchmarking arguments:

    <variablelist>
     <varlistentry>
      <term><option>-b</> <replaceable>scriptname[@weight]</></term>
      <term><option>--builtin</>=<replaceable>scriptname[@weight]</></term>
      <listitem>
       <para>
        Add the specified built-in script to the list of executed scripts.
        An optional integer weight after <literal>@</> allows to adjust the
        probability of drawing the script.  If not specified, it is set to 1.
        Available built-in scripts are: <literal>tpcb-like</>,
        <literal>simple-update</> and <literal>select-only</>.
        Unambiguous prefixes of built-in names are accepted.
        With special name <literal>list</>, show the list of built-in scripts
        and exit immediately.
       </para>
      </listitem>
     </varlistentry>


     <varlistentry>
      <term><option>-c</option> <replaceable>clients</></term>
      <term><option>--client=</option><replaceable>clients</></term>
      <listitem>
       <para>
        Number of clients simulated, that is, number of concurrent database
        sessions.  Default is 1.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-C</option></term>
      <term><option>--connect</option></term>
      <listitem>
       <para>
        Establish a new connection for each transaction, rather than
        doing it just once per client session.
        This is useful to measure the connection overhead.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-d</option></term>
      <term><option>--debug</option></term>
      <listitem>
       <para>
        Print debugging output.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-D</option> <replaceable>varname</><literal>=</><replaceable>value</></term>
      <term><option>--define=</option><replaceable>varname</><literal>=</><replaceable>value</></term>
      <listitem>
       <para>
        Define a variable for use by a custom script (see below).
        Multiple <option>-D</> options are allowed.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-f</> <replaceable>filename[@weight]</></term>
      <term><option>--file=</><replaceable>filename[@weight]</></term>
      <listitem>
       <para>
        Add a transaction script read from <replaceable>filename</> to
        the list of executed scripts.
        An optional integer weight after <literal>@</> allows to adjust the
        probability of drawing the test.
        See below for details.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-j</option> <replaceable>threads</></term>
      <term><option>--jobs=</option><replaceable>threads</></term>
      <listitem>
       <para>
        Number of worker threads within <application>pgbench</application>.
        Using more than one thread can be helpful on multi-CPU machines.
        Clients are distributed as evenly as possible among available threads.
        Default is 1.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-l</option></term>
      <term><option>--log</option></term>
      <listitem>
       <para>
        Write information about each transaction to a log file.
        See below for details.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-L</option> <replaceable>limit</></term>
      <term><option>--latency-limit=</option><replaceable>limit</></term>
      <listitem>
       <para>
        Transaction which last more than <replaceable>limit</> milliseconds
        are counted and reported separately, as <firstterm>late</>.
       </para>
       <para>
        When throttling is used (<option>--rate=...</>), transactions that
        lag behind schedule by more than <replaceable>limit</> ms, and thus
        have no hope of meeting the latency limit, are not sent to the server
        at all. They are counted and reported separately as
        <firstterm>skipped</>.
       </para>
       </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-M</option> <replaceable>querymode</></term>
      <term><option>--protocol=</option><replaceable>querymode</></term>
      <listitem>
       <para>
        Protocol to use for submitting queries to the server:
          <itemizedlist>
           <listitem>
            <para><literal>simple</>: use simple query protocol.</para>
           </listitem>
           <listitem>
            <para><literal>extended</>: use extended query protocol.</para>
           </listitem>
           <listitem>
            <para><literal>prepared</>: use extended query protocol with prepared statements.</para>
           </listitem>
          </itemizedlist>
        The default is simple query protocol.  (See <xref linkend="protocol">
        for more information.)
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-n</option></term>
      <term><option>--no-vacuum</option></term>
      <listitem>
       <para>
        Perform no vacuuming before running the test.
        This option is <emphasis>necessary</>
        if you are running a custom test scenario that does not include
        the standard tables <structname>pgbench_accounts</>,
        <structname>pgbench_branches</>, <structname>pgbench_history</>, and
        <structname>pgbench_tellers</>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-N</option></term>
      <term><option>--skip-some-updates</option></term>
      <listitem>
       <para>
        Run built-in simple-update script.
        Shorthand for <option>-b simple-update</>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-P</option> <replaceable>sec</></term>
      <term><option>--progress=</option><replaceable>sec</></term>
      <listitem>
       <para>
        Show progress report every <replaceable>sec</> seconds.  The report
        includes the time since the beginning of the run, the tps since the
        last report, and the transaction latency average and standard
        deviation since the last report.  Under throttling (<option>-R</>),
        the latency is computed with respect to the transaction scheduled
        start time, not the actual transaction beginning time, thus it also
        includes the average schedule lag time.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-r</option></term>
      <term><option>--report-latencies</option></term>
      <listitem>
       <para>
        Report the average per-statement latency (execution time from the
        perspective of the client) of each command after the benchmark
        finishes.  See below for details.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-R</option> <replaceable>rate</></term>
      <term><option>--rate=</option><replaceable>rate</></term>
      <listitem>
       <para>
        Execute transactions targeting the specified rate instead of running
        as fast as possible (the default).  The rate is given in transactions
        per second.  If the targeted rate is above the maximum possible rate,
        the rate limit won't impact the results.
       </para>
       <para>
        The rate is targeted by starting transactions along a
        Poisson-distributed schedule time line.  The expected start time
        schedule moves forward based on when the client first started, not
        when the previous transaction ended.  That approach means that when
        transactions go past their original scheduled end time, it is
        possible for later ones to catch up again.
       </para>
       <para>
        When throttling is active, the transaction latency reported at the
        end of the run is calculated from the scheduled start times, so it
        includes the time each transaction had to wait for the previous
        transaction to finish. The wait time is called the schedule lag time,
        and its average and maximum are also reported separately. The
        transaction latency with respect to the actual transaction start time,
        i.e. the time spent executing the transaction in the database, can be
        computed by subtracting the schedule lag time from the reported
        latency.
       </para>

       <para>
        If <option>--latency-limit</> is used together with <option>--rate</>,
        a transaction can lag behind so much that it is already over the
        latency limit when the previous transaction ends, because the latency
        is calculated from the scheduled start time. Such transactions are
        not sent to the server, but are skipped altogether and counted
        separately.
       </para>

       <para>
        A high schedule lag time is an indication that the system cannot
        process transactions at the specified rate, with the chosen number of
        clients and threads. When the average transaction execution time is
        longer than the scheduled interval between each transaction, each
        successive transaction will fall further behind, and the schedule lag
        time will keep increasing the longer the test run is. When that
        happens, you will have to reduce the specified transaction rate.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-s</option> <replaceable>scale_factor</></term>
      <term><option>--scale=</option><replaceable>scale_factor</></term>
      <listitem>
       <para>
        Report the specified scale factor in <application>pgbench</>'s
        output.  With the built-in tests, this is not necessary; the
        correct scale factor will be detected by counting the number of
        rows in the <structname>pgbench_branches</> table.
        However, when testing only custom benchmarks (<option>-f</> option),
        the scale factor will be reported as 1 unless this option is used.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-S</option></term>
      <term><option>--select-only</option></term>
      <listitem>
       <para>
        Run built-in select-only script.
        Shorthand for <option>-b select-only</>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-t</option> <replaceable>transactions</></term>
      <term><option>--transactions=</option><replaceable>transactions</></term>
      <listitem>
       <para>
        Number of transactions each client runs.  Default is 10.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-T</option> <replaceable>seconds</></term>
      <term><option>--time=</option><replaceable>seconds</></term>
      <listitem>
       <para>
        Run the test for this many seconds, rather than a fixed number of
        transactions per client. <option>-t</option> and
        <option>-T</option> are mutually exclusive.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-v</option></term>
      <term><option>--vacuum-all</option></term>
      <listitem>
       <para>
        Vacuum all four standard tables before running the test.
        With neither <option>-n</> nor <option>-v</>, <application>pgbench</application> will vacuum the
        <structname>pgbench_tellers</> and <structname>pgbench_branches</>
        tables, and will truncate <structname>pgbench_history</>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--aggregate-interval=<replaceable>seconds</></option></term>
      <listitem>
       <para>
        Length of aggregation interval (in seconds).  May be used only
        with <option>-l</option> option.  With this option, the log contains
        per-interval summary data, as described below.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--log-prefix=<replaceable>prefix</></option></term>
      <listitem>
       <para>
        Set the filename prefix for the log files created by
        <option>--log</>.  The default is <literal>pgbench_log</>.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--progress-timestamp</option></term>
      <listitem>
       <para>
        When showing progress (option <option>-P</>), use a timestamp
        (Unix epoch) instead of the number of seconds since the
        beginning of the run.  The unit is in seconds, with millisecond
        precision after the dot.
        This helps compare logs generated by various tools.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>--sampling-rate=<replaceable>rate</></option></term>
      <listitem>
       <para>
        Sampling rate, used when writing data into the log, to reduce the
        amount of log generated. If this option is given, only the specified
        fraction of transactions are logged. 1.0 means all transactions will
        be logged, 0.05 means only 5% of the transactions will be logged.
       </para>
       <para>
        Remember to take the sampling rate into account when processing the
        log file. For example, when computing tps values, you need to multiply
        the numbers accordingly (e.g. with 0.01 sample rate, you'll only get
        1/100 of the actual tps).
       </para>
      </listitem>
     </varlistentry>

    </variablelist>
   </para>

 </refsect2>

 <refsect2 id="pgbench-common-options">
  <title>Common Options</title>

   <para>
    <application>pgbench</application> accepts the following command-line
    common arguments:

    <variablelist>

     <varlistentry>
      <term><option>-h</option> <replaceable>hostname</></term>
      <term><option>--host=</option><replaceable>hostname</></term>
      <listitem>
       <para>
        The database server's host name
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-p</option> <replaceable>port</></term>
      <term><option>--port=</option><replaceable>port</></term>
      <listitem>
       <para>
        The database server's port number
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-U</option> <replaceable>login</></term>
      <term><option>--username=</option><replaceable>login</></term>
      <listitem>
       <para>
        The user name to connect as
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-V</></term>
      <term><option>--version</></term>
      <listitem>
       <para>
        Print the <application>pgbench</application> version and exit.
       </para>
      </listitem>
     </varlistentry>

     <varlistentry>
      <term><option>-?</></term>
      <term><option>--help</></term>
      <listitem>
       <para>
        Show help about <application>pgbench</application> command line
        arguments, and exit.
       </para>
      </listitem>
     </varlistentry>
    </variablelist>
   </para>

 </refsect2>
 </refsect1>

 <refsect1>
  <title>Notes</title>

 <refsect2>
  <title>What is the <quote>Transaction</> Actually Performed in <application>pgbench</application>?</title>

  <para>
   <application>pgbench</> executes test scripts chosen randomly
   from a specified list.
   They include built-in scripts with <option>-b</> and
   user-provided custom scripts with <option>-f</>.
   Each script may be given a relative weight specified after a
   <literal>@</> so as to change its drawing probability.
   The default weight is <literal>1</>.
   Scripts with a weight of <literal>0</> are ignored.
 </para>

  <para>
   The default built-in transaction script (also invoked with <option>-b tpcb-like</>)
   issues seven commands per transaction over randomly chosen <literal>aid</>,
   <literal>tid</>, <literal>bid</> and <literal>balance</>.
   The scenario is inspired by the TPC-B benchmark, but is not actually TPC-B,
   hence the name.
  </para>

  <orderedlist>
   <listitem><para><literal>BEGIN;</literal></para></listitem>
   <listitem><para><literal>UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;</literal></para></listitem>
   <listitem><para><literal>SELECT abalance FROM pgbench_accounts WHERE aid = :aid;</literal></para></listitem>
   <listitem><para><literal>UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;</literal></para></listitem>
   <listitem><para><literal>UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;</literal></para></listitem>
   <listitem><para><literal>INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);</literal></para></listitem>
   <listitem><para><literal>END;</literal></para></listitem>
  </orderedlist>

  <para>
   If you select the <literal>simple-update</> built-in (also <option>-N</>),
   steps 4 and 5 aren't included in the transaction.
   This will avoid update contention on these tables, but
   it makes the test case even less like TPC-B.
  </para>

  <para>
   If you select the <literal>select-only</> built-in (also <option>-S</>),
   only the <command>SELECT</> is issued.
  </para>
 </refsect2>

 <refsect2>
  <title>Custom Scripts</title>

  <para>
   <application>pgbench</application> has support for running custom
   benchmark scenarios by replacing the default transaction script
   (described above) with a transaction script read from a file
   (<option>-f</option> option).  In this case a <quote>transaction</>
   counts as one execution of a script file.
  </para>

  <para>
   A script file contains one or more SQL commands terminated by
   semicolons.  Empty lines and lines beginning with
   <literal>--</> are ignored.  Script files can also contain
   <quote>meta commands</>, which are interpreted by <application>pgbench</>
   itself, as described below.
  </para>

  <note>
   <para>
    Before <productname>PostgreSQL</> 9.6, SQL commands in script files
    were terminated by newlines, and so they could not be continued across
    lines.  Now a semicolon is <emphasis>required</> to separate consecutive
    SQL commands (though a SQL command does not need one if it is followed
    by a meta command).  If you need to create a script file that works with
    both old and new versions of <application>pgbench</>, be sure to write
    each SQL command on a single line ending with a semicolon.
   </para>
  </note>

  <para>
   There is a simple variable-substitution facility for script files.
   Variables can be set by the command-line <option>-D</> option,
   explained above, or by the meta commands explained below.
   In addition to any variables preset by <option>-D</> command-line options,
   there are a few variables that are preset automatically, listed in
   <xref linkend="pgbench-automatic-variables">. A value specified for these
   variables using <option>-D</> takes precedence over the automatic presets.
   Once set, a variable's
   value can be inserted into a SQL command by writing
   <literal>:</><replaceable>variablename</>.  When running more than
   one client session, each session has its own set of variables.
  </para>

   <table id="pgbench-automatic-variables">
    <title>Automatic Variables</title>
    <tgroup cols="2">
     <thead>
      <row>
       <entry>Variable</entry>
       <entry>Description</entry>
      </row>
     </thead>

     <tbody>
      <row>
       <entry> <literal>scale</literal> </entry>
       <entry>current scale factor</entry>
      </row>

      <row>
       <entry> <literal>client_id</literal> </entry>
       <entry>unique number identifying the client session (starts from zero)</entry>
      </row>
     </tbody>
    </tgroup>
   </table>

  <para>
   Script file meta commands begin with a backslash (<literal>\</>) and
   normally extend to the end of the line, although they can be continued
   to additional lines by writing backslash-return.
   Arguments to a meta command are separated by white space.
   These meta commands are supported:
  </para>

  <variablelist>
   <varlistentry id='pgbench-metacommand-set'>
    <term>
     <literal>\set <replaceable>varname</> <replaceable>expression</></literal>
    </term>

    <listitem>
     <para>
      Sets variable <replaceable>varname</> to a value calculated
      from <replaceable>expression</>.
      The expression may contain integer constants such as <literal>5432</>,
      double constants such as <literal>3.14159</>,
      references to variables <literal>:</><replaceable>variablename</>,
      unary operators (<literal>+</>, <literal>-</>) and binary operators
      (<literal>+</>, <literal>-</>, <literal>*</>, <literal>/</>,
      <literal>%</>) with their usual precedence and associativity,
      <link linkend="pgbench-builtin-functions">function calls</>, and
      parentheses.
     </para>

     <para>
      Examples:
<programlisting>
\set ntellers 10 * :scale
\set aid (1021 * random(1, 100000 * :scale)) % \
           (100000 * :scale) + 1
</programlisting></para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term>
     <literal>\sleep <replaceable>number</> [ us | ms | s ]</literal>
    </term>

    <listitem>
     <para>
      Causes script execution to sleep for the specified duration in
      microseconds (<literal>us</>), milliseconds (<literal>ms</>) or seconds
      (<literal>s</>).  If the unit is omitted then seconds are the default.
      <replaceable>number</> can be either an integer constant or a
      <literal>:</><replaceable>variablename</> reference to a variable
      having an integer value.
     </para>

     <para>
      Example:
<programlisting>
\sleep 10 ms
</programlisting></para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term>
     <literal>\setshell <replaceable>varname</> <replaceable>command</> [ <replaceable>argument</> ... ]</literal>
    </term>

    <listitem>
     <para>
      Sets variable <replaceable>varname</> to the result of the shell command
      <replaceable>command</> with the given <replaceable>argument</>(s).
      The command must return an integer value through its standard output.
     </para>

     <para>
      <replaceable>command</> and each <replaceable>argument</> can be either
      a text constant or a <literal>:</><replaceable>variablename</> reference
      to a variable. If you want to use an <replaceable>argument</> starting
      with a colon, write an additional colon at the beginning of
      <replaceable>argument</>.
     </para>

     <para>
      Example:
<programlisting>
\setshell variable_to_be_assigned command literal_argument :variable ::literal_starting_with_colon
</programlisting></para>
    </listitem>
   </varlistentry>

   <varlistentry>
    <term>
     <literal>\shell <replaceable>command</> [ <replaceable>argument</> ... ]</literal>
    </term>

    <listitem>
     <para>
      Same as <literal>\setshell</literal>, but the result of the command
      is discarded.
     </para>

     <para>
      Example:
<programlisting>
\shell command literal_argument :variable ::literal_starting_with_colon
</programlisting></para>
    </listitem>
   </varlistentry>
  </variablelist>
 </refsect2>

 <refsect2 id="pgbench-builtin-functions">
  <title>Built-In Functions</title>

  <para>
   The functions listed in <xref linkend="pgbench-functions"> are built
   into <application>pgbench</> and may be used in expressions appearing in
   <link linkend="pgbench-metacommand-set"><literal>\set</literal></link>.
  </para>

   <!-- list pgbench functions in alphabetical order -->
   <table id="pgbench-functions">
    <title>pgbench Functions</title>
    <tgroup cols="5">
     <thead>
      <row>
       <entry>Function</entry>
       <entry>Return Type</entry>
       <entry>Description</entry>
       <entry>Example</entry>
       <entry>Result</entry>
      </row>
     </thead>
     <tbody>
      <row>
       <entry><literal><function>abs(<replaceable>a</>)</></></>
       <entry>same as <replaceable>a</></>
       <entry>absolute value</>
       <entry><literal>abs(-17)</></>
       <entry><literal>17</></>
      </row>
      <row>
       <entry><literal><function>debug(<replaceable>a</>)</></></>
       <entry>same as <replaceable>a</> </>
       <entry>print <replaceable>a</> to <systemitem>stderr</systemitem>,
        and return <replaceable>a</></>
       <entry><literal>debug(5432.1)</></>
       <entry><literal>5432.1</></>
      </row>
      <row>
       <entry><literal><function>double(<replaceable>i</>)</></></>
       <entry>double</>
       <entry>cast to double</>
       <entry><literal>double(5432)</></>
       <entry><literal>5432.0</></>
      </row>
      <row>
       <entry><literal><function>greatest(<replaceable>a</> [, <replaceable>...</> ] )</></></>
       <entry>double if any <replaceable>a</> is double, else integer</>
       <entry>largest value among arguments</>
       <entry><literal>greatest(5, 4, 3, 2)</></>
       <entry><literal>5</></>
      </row>
      <row>
       <entry><literal><function>int(<replaceable>x</>)</></></>
       <entry>integer</>
       <entry>cast to int</>
       <entry><literal>int(5.4 + 3.8)</></>
       <entry><literal>9</></>
      </row>
      <row>
       <entry><literal><function>least(<replaceable>a</> [, <replaceable>...</> ] )</></></>
       <entry>double if any <replaceable>a</> is double, else integer</>
       <entry>smallest value among arguments</>
       <entry><literal>least(5, 4, 3, 2.1)</></>
       <entry><literal>2.1</></>
      </row>
      <row>
       <entry><literal><function>pi()</></></>
       <entry>double</>
       <entry>value of the constant PI</>
       <entry><literal>pi()</></>
       <entry><literal>3.14159265358979323846</></>
      </row>
      <row>
       <entry><literal><function>random(<replaceable>lb</>, <replaceable>ub</>)</></></>
       <entry>integer</>
       <entry>uniformly-distributed random integer in <literal>[lb, ub]</></>
       <entry><literal>random(1, 10)</></>
       <entry>an integer between <literal>1</> and <literal>10</></>
      </row>
      <row>
       <entry><literal><function>random_exponential(<replaceable>lb</>, <replaceable>ub</>, <replaceable>parameter</>)</></></>
       <entry>integer</>
       <entry>exponentially-distributed random integer in <literal>[lb, ub]</>,
              see below</>
       <entry><literal>random_exponential(1, 10, 3.0)</></>
       <entry>an integer between <literal>1</> and <literal>10</></>
      </row>
      <row>
       <entry><literal><function>random_gaussian(<replaceable>lb</>, <replaceable>ub</>, <replaceable>parameter</>)</></></>
       <entry>integer</>
       <entry>Gaussian-distributed random integer in <literal>[lb, ub]</>,
              see below</>
       <entry><literal>random_gaussian(1, 10, 2.5)</></>
       <entry>an integer between <literal>1</> and <literal>10</></>
      </row>
      <row>
       <entry><literal><function>sqrt(<replaceable>x</>)</></></>
       <entry>double</>
       <entry>square root</>
       <entry><literal>sqrt(2.0)</></>
       <entry><literal>1.414213562</></>
      </row>
     </tbody>
     </tgroup>
   </table>

   <para>
    The <literal>random</> function generates values using a uniform
    distribution, that is all the values are drawn within the specified
    range with equal probability. The <literal>random_exponential</> and
    <literal>random_gaussian</> functions require an additional double
    parameter which determines the precise shape of the distribution.
   </para>

   <itemizedlist>
    <listitem>
     <para>
      For an exponential distribution, <replaceable>parameter</>
      controls the distribution by truncating a quickly-decreasing
      exponential distribution at <replaceable>parameter</>, and then
      projecting onto integers between the bounds.
      To be precise, with
<literallayout>
f(x) = exp(-parameter * (x - min) / (max - min + 1)) / (1 - exp(-parameter))
</literallayout>
      Then value <replaceable>i</> between <replaceable>min</> and
      <replaceable>max</> inclusive is drawn with probability:
      <literal>f(x) - f(x + 1)</>.
     </para>

     <para>
      Intuitively, the larger the <replaceable>parameter</>, the more
      frequently values close to <replaceable>min</> are accessed, and the
      less frequently values close to <replaceable>max</> are accessed.
      The closer to 0 <replaceable>parameter</> is, the flatter (more
      uniform) the access distribution.
      A crude approximation of the distribution is that the most frequent 1%
      values in the range, close to <replaceable>min</>, are drawn
      <replaceable>parameter</>% of the time.
      The <replaceable>parameter</> value must be strictly positive.
     </para>
    </listitem>

    <listitem>
     <para>
      For a Gaussian distribution, the interval is mapped onto a standard
      normal distribution (the classical bell-shaped Gaussian curve) truncated
      at <literal>-parameter</> on the left and <literal>+parameter</>
      on the right.
      Values in the middle of the interval are more likely to be drawn.
      To be precise, if <literal>PHI(x)</> is the cumulative distribution
      function of the standard normal distribution, with mean <literal>mu</>
      defined as <literal>(max + min) / 2.0</>, with
<literallayout>
f(x) = PHI(2.0 * parameter * (x - mu) / (max - min + 1)) /
       (2.0 * PHI(parameter) - 1)
</literallayout>
      then value <replaceable>i</> between <replaceable>min</> and
      <replaceable>max</> inclusive is drawn with probability:
      <literal>f(i + 0.5) - f(i - 0.5)</>.
      Intuitively, the larger the <replaceable>parameter</>, the more
      frequently values close to the middle of the interval are drawn, and the
      less frequently values close to the <replaceable>min</> and
      <replaceable>max</> bounds. About 67% of values are drawn from the
      middle <literal>1.0 / parameter</>, that is a relative
      <literal>0.5 / parameter</> around the mean, and 95% in the middle
      <literal>2.0 / parameter</>, that is a relative
      <literal>1.0 / parameter</> around the mean; for instance, if
      <replaceable>parameter</> is 4.0, 67% of values are drawn from the
      middle quarter (1.0 / 4.0) of the interval (i.e. from
      <literal>3.0 / 8.0</> to <literal>5.0 / 8.0</>) and 95% from
      the middle half (<literal>2.0 / 4.0</>) of the interval (second and third
      quartiles). The minimum <replaceable>parameter</> is 2.0 for performance
      of the Box-Muller transform.
     </para>
    </listitem>
   </itemizedlist>

  <para>
   As an example, the full definition of the built-in TPC-B-like
   transaction is:

<programlisting>
\set aid random(1, 100000 * :scale)
\set bid random(1, 1 * :scale)
\set tid random(1, 10 * :scale)
\set delta random(-5000, 5000)
BEGIN;
UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
END;
</programlisting>

   This script allows each iteration of the transaction to reference
   different, randomly-chosen rows.  (This example also shows why it's
   important for each client session to have its own variables &mdash;
   otherwise they'd not be independently touching different rows.)
  </para>

 </refsect2>

 <refsect2>
  <title>Per-Transaction Logging</title>

  <para>
   With the <option>-l</> option (but without
   the <option>--aggregate-interval</option> option),
   <application>pgbench</> writes information about each transaction
   to a log file.  The log file will be named
   <filename><replaceable>prefix</>.<replaceable>nnn</></filename>,
   where <replaceable>prefix</> defaults to <literal>pgbench_log</>, and
   <replaceable>nnn</> is the PID of the
   <application>pgbench</application> process.
   The prefix can be changed by using the <option>--log-prefix</> option.
   If the <option>-j</> option is 2 or higher, so that there are multiple
   worker threads, each will have its own log file. The first worker will
   use the same name for its log file as in the standard single worker case.
   The additional log files for the other workers will be named
   <filename><replaceable>prefix</>.<replaceable>nnn</>.<replaceable>mmm</></filename>,
   where <replaceable>mmm</> is a sequential number for each worker starting
   with 1.
  </para>

  <para>
   The format of the log is:

<synopsis>
<replaceable>client_id</> <replaceable>transaction_no</> <replaceable>time</> <replaceable>script_no</> <replaceable>time_epoch</> <replaceable>time_us</> <optional> <replaceable>schedule_lag</replaceable> </optional>
</synopsis>

   where
   <replaceable>client_id</> indicates which client session ran the transaction,
   <replaceable>transaction_no</> counts how many transactions have been
   run by that session,
   <replaceable>time</> is the total elapsed transaction time in microseconds,
   <replaceable>script_no</> identifies which script file was used (useful when
   multiple scripts were specified with <option>-f</> or <option>-b</>),
   and <replaceable>time_epoch</>/<replaceable>time_us</> are a
   Unix-epoch time stamp and an offset
   in microseconds (suitable for creating an ISO 8601
   time stamp with fractional seconds) showing when
   the transaction completed.
   The <replaceable>schedule_lag</> field is the difference between the
   transaction's scheduled start time, and the time it actually started, in
   microseconds. It is only present when the <option>--rate</> option is used.
   When both <option>--rate</> and <option>--latency-limit</> are used,
   the <replaceable>time</> for a skipped transaction will be reported as
   <literal>skipped</>.
  </para>

  <para>
   Here is a snippet of a log file generated in a single-client run:
<screen>
0 199 2241 0 1175850568 995598
0 200 2465 0 1175850568 998079
0 201 2513 0 1175850569 608
0 202 2038 0 1175850569 2663
</screen>

   Another example with <literal>--rate=100</>
   and <literal>--latency-limit=5</> (note the additional
   <replaceable>schedule_lag</> column):
<screen>
0 81 4621 0 1412881037 912698 3005
0 82 6173 0 1412881037 914578 4304
0 83 skipped 0 1412881037 914578 5217
0 83 skipped 0 1412881037 914578 5099
0 83 4722 0 1412881037 916203 3108
0 84 4142 0 1412881037 918023 2333
0 85 2465 0 1412881037 919759 740
</screen>
   In this example, transaction 82 was late, because its latency (6.173 ms) was
   over the 5 ms limit. The next two transactions were skipped, because they
   were already late before they were even started.
  </para>

  <para>
   When running a long test on hardware that can handle a lot of transactions,
   the log files can become very large.  The <option>--sampling-rate</> option
   can be used to log only a random sample of transactions.
  </para>
 </refsect2>

 <refsect2>
  <title>Aggregated Logging</title>

  <para>
   With the <option>--aggregate-interval</option> option, a different
   format is used for the log files:

<synopsis>
<replaceable>interval_start</> <replaceable>num_transactions</> <replaceable>sum_latency</> <replaceable>sum_latency_2</> <replaceable>min_latency</> <replaceable>max_latency</> <optional> <replaceable>sum_lag</> <replaceable>sum_lag_2</> <replaceable>min_lag</> <replaceable>max_lag</> <optional> <replaceable>skipped</> </optional> </optional>
</synopsis>

   where
   <replaceable>interval_start</> is the start of the interval (as a Unix
   epoch time stamp),
   <replaceable>num_transactions</> is the number of transactions
   within the interval,
   <replaceable>sum_latency</replaceable> is the sum of the transaction
   latencies within the interval,
   <replaceable>sum_latency_2</replaceable> is the sum of squares of the
   transaction latencies within the interval,
   <replaceable>min_latency</> is the minimum latency within the interval,
   and
   <replaceable>max_latency</> is the maximum latency within the interval.
   The next fields,
   <replaceable>sum_lag</>, <replaceable>sum_lag_2</>, <replaceable>min_lag</>,
   and <replaceable>max_lag</>, are only present if the <option>--rate</>
   option is used.
   They provide statistics about the time each transaction had to wait for the
   previous one to finish, i.e. the difference between each transaction's
   scheduled start time and the time it actually started.
   The very last field, <replaceable>skipped</>,
   is only present if the <option>--latency-limit</> option is used, too.
   It counts the number of transactions skipped because they would have
   started too late.
   Each transaction is counted in the interval when it was committed.
  </para>

  <para>
   Here is some example output:
<screen>
1345828501 5601 1542744 483552416 61 2573
1345828503 7884 1979812 565806736 60 1479
1345828505 7208 1979422 567277552 59 1391
1345828507 7685 1980268 569784714 60 1398
1345828509 7073 1979779 573489941 236 1411
</screen></para>

  <para>
   Notice that while the plain (unaggregated) log file shows which script
   was used for each transaction, the aggregated log does not. Therefore if
   you need per-script data, you need to aggregate the data on your own.
  </para>

 </refsect2>

 <refsect2>
  <title>Per-Statement Latencies</title>

  <para>
   With the <option>-r</> option, <application>pgbench</> collects
   the elapsed transaction time of each statement executed by every
   client.  It then reports an average of those values, referred to
   as the latency for each statement, after the benchmark has finished.
  </para>

  <para>
   For the default script, the output will look similar to this:
<screen>
starting vacuum...end.
transaction type: &lt;builtin: TPC-B (sort of)&gt;
scaling factor: 1
query mode: simple
number of clients: 10
number of threads: 1
number of transactions per client: 1000
number of transactions actually processed: 10000/10000
latency average = 15.844 ms
latency stddev = 2.715 ms
tps = 618.764555 (including connections establishing)
tps = 622.977698 (excluding connections establishing)
script statistics:
 - statement latencies in milliseconds:
        0.002  \set aid random(1, 100000 * :scale)
        0.005  \set bid random(1, 1 * :scale)
        0.002  \set tid random(1, 10 * :scale)
        0.001  \set delta random(-5000, 5000)
        0.326  BEGIN;
        0.603  UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
        0.454  SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
        5.528  UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
        7.335  UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
        0.371  INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
        1.212  END;
</screen>
  </para>

  <para>
   If multiple script files are specified, the averages are reported
   separately for each script file.
  </para>

  <para>
   Note that collecting the additional timing information needed for
   per-statement latency computation adds some overhead.  This will slow
   average execution speed and lower the computed TPS.  The amount
   of slowdown varies significantly depending on platform and hardware.
   Comparing average TPS values with and without latency reporting enabled
   is a good way to measure if the timing overhead is significant.
  </para>
 </refsect2>

 <refsect2>
  <title>Good Practices</title>

  <para>
   It is very easy to use <application>pgbench</> to produce completely
   meaningless numbers.  Here are some guidelines to help you get useful
   results.
  </para>

  <para>
   In the first place, <emphasis>never</> believe any test that runs
   for only a few seconds.  Use the <option>-t</> or <option>-T</> option
   to make the run last at least a few minutes, so as to average out noise.
   In some cases you could need hours to get numbers that are reproducible.
   It's a good idea to try the test run a few times, to find out if your
   numbers are reproducible or not.
  </para>

  <para>
   For the default TPC-B-like test scenario, the initialization scale factor
   (<option>-s</>) should be at least as large as the largest number of
   clients you intend to test (<option>-c</>); else you'll mostly be
   measuring update contention.  There are only <option>-s</> rows in
   the <structname>pgbench_branches</> table, and every transaction wants to
   update one of them, so <option>-c</> values in excess of <option>-s</>
   will undoubtedly result in lots of transactions blocked waiting for
   other transactions.
  </para>

  <para>
   The default test scenario is also quite sensitive to how long it's been
   since the tables were initialized: accumulation of dead rows and dead space
   in the tables changes the results.  To understand the results you must keep
   track of the total number of updates and when vacuuming happens.  If
   autovacuum is enabled it can result in unpredictable changes in measured
   performance.
  </para>

  <para>
   A limitation of <application>pgbench</> is that it can itself become
   the bottleneck when trying to test a large number of client sessions.
   This can be alleviated by running <application>pgbench</> on a different
   machine from the database server, although low network latency will be
   essential.  It might even be useful to run several <application>pgbench</>
   instances concurrently, on several client machines, against the same
   database server.
  </para>
 </refsect2>
 </refsect1>
</refentry>