1PGBENCH(1) PostgreSQL 12.6 Documentation PGBENCH(1)
2
6 pgbench - run a benchmark test on PostgreSQL
7
9 pgbench -i [option...] [dbname]
10 pgbench [option...] [dbname]
12
14 pgbench is a simple program for running benchmark tests on PostgreSQL.
15 It runs the same sequence of SQL commands over and over, possibly in
16 multiple concurrent database sessions, and then calculates the average
17 transaction rate (transactions per second). By default, pgbench tests a
18 scenario that is loosely based on TPC-B, involving five SELECT, UPDATE,
19 and INSERT commands per transaction. However, it is easy to test other
20 cases by writing your own transaction script files.
21 Typical output from pgbench looks like:
23 transaction type: <builtin: TPC-B (sort of)>
25 scaling factor: 10
26 query mode: simple
27 number of clients: 10
28 number of threads: 1
29 number of transactions per client: 1000
30 number of transactions actually processed: 10000/10000
31 tps = 85.184871 (including connections establishing)
32 tps = 85.296346 (excluding connections establishing)
33 The first six lines report some of the most important parameter
35 settings. The next line reports the number of transactions completed
36 and intended (the latter being just the product of number of clients
37 and number of transactions per client); these will be equal unless the
38 run failed before completion. (In -T mode, only the actual number of
39 transactions is printed.) The last two lines report the number of
40 transactions per second, figured with and without counting the time to
41 start database sessions.
42 The default TPC-B-like transaction test requires specific tables to be
44 set up beforehand. pgbench should be invoked with the -i (initialize)
45 option to create and populate these tables. (When you are testing a
46 custom script, you don't need this step, but will instead need to do
47 whatever setup your test needs.) Initialization looks like:
48 pgbench -i [ other-options ] dbname
50 where dbname is the name of the already-created database to test in.
52 (You may also need -h, -p, and/or -U options to specify how to connect
53 to the database server.)
54 Caution
56 pgbench -i creates four tables pgbench_accounts, pgbench_branches,
57 pgbench_history, and pgbench_tellers, destroying any existing
58 tables of these names. Be very careful to use another database if
59 you have tables having these names!
60 At the default “scale factor” of 1, the tables initially contain this
62 many rows:
63 table # of rows
65 ---------------------------------
66 pgbench_branches 1
67 pgbench_tellers 10
68 pgbench_accounts 100000
69 pgbench_history 0
70 You can (and, for most purposes, probably should) increase the number
72 of rows by using the -s (scale factor) option. The -F (fillfactor)
73 option might also be used at this point.
74 Once you have done the necessary setup, you can run your benchmark with
76 a command that doesn't include -i, that is
77 pgbench [ options ] dbname
79 In nearly all cases, you'll need some options to make a useful test.
81 The most important options are -c (number of clients), -t (number of
82 transactions), -T (time limit), and -f (specify a custom script file).
83 See below for a full list.
84
86 The following is divided into three subsections. Different options are
87 used during database initialization and while running benchmarks, but
88 some options are useful in both cases.
89 Initialization Options
91 pgbench accepts the following command-line initialization arguments:
92 -i
94 --initialize
95 Required to invoke initialization mode.
96 -I init_steps
98 --init-steps=init_steps
99 Perform just a selected set of the normal initialization steps.
100 init_steps specifies the initialization steps to be performed,
101 using one character per step. Each step is invoked in the specified
102 order. The default is dtgvp. The available steps are:
103 d (Drop)
105 Drop any existing pgbench tables.
106 t (create Tables)
108 Create the tables used by the standard pgbench scenario, namely
109 pgbench_accounts, pgbench_branches, pgbench_history, and
110 pgbench_tellers.
111 g (Generate data)
113 Generate data and load it into the standard tables, replacing
114 any data already present.
115 v (Vacuum)
117 Invoke VACUUM on the standard tables.
118 p (create Primary keys)
120 Create primary key indexes on the standard tables.
121 f (create Foreign keys)
123 Create foreign key constraints between the standard tables.
124 (Note that this step is not performed by default.)
125 -F fillfactor
128 --fillfactor=fillfactor
129 Create the pgbench_accounts, pgbench_tellers and pgbench_branches
130 tables with the given fillfactor. Default is 100.
131 -n
133 --no-vacuum
134 Perform no vacuuming during initialization. (This option suppresses
135 the v initialization step, even if it was specified in -I.)
136 -q
138 --quiet
139 Switch logging to quiet mode, producing only one progress message
140 per 5 seconds. The default logging prints one message each 100000
141 rows, which often outputs many lines per second (especially on good
142 hardware).
143 -s scale_factor
145 --scale=scale_factor
146 Multiply the number of rows generated by the scale factor. For
147 example, -s 100 will create 10,000,000 rows in the pgbench_accounts
148 table. Default is 1. When the scale is 20,000 or larger, the
149 columns used to hold account identifiers (aid columns) will switch
150 to using larger integers (bigint), in order to be big enough to
151 hold the range of account identifiers.
152 --foreign-keys
154 Create foreign key constraints between the standard tables. (This
155 option adds the f step to the initialization step sequence, if it
156 is not already present.)
157 --index-tablespace=index_tablespace
159 Create indexes in the specified tablespace, rather than the default
160 tablespace.
161 --tablespace=tablespace
163 Create tables in the specified tablespace, rather than the default
164 tablespace.
165 --unlogged-tables
167 Create all tables as unlogged tables, rather than permanent tables.
168 Benchmarking Options
170 pgbench accepts the following command-line benchmarking arguments:
171 -b scriptname[@weight]
173 --builtin=scriptname[@weight]
174 Add the specified built-in script to the list of scripts to be
175 executed. Available built-in scripts are: tpcb-like, simple-update
176 and select-only. Unambiguous prefixes of built-in names are
177 accepted. With the special name list, show the list of built-in
178 scripts and exit immediately.
179 Optionally, write an integer weight after @ to adjust the
181 probability of selecting this script versus other ones. The default
182 weight is 1. See below for details.
183 -c clients
185 --client=clients
186 Number of clients simulated, that is, number of concurrent database
187 sessions. Default is 1.
188 -C
190 --connect
191 Establish a new connection for each transaction, rather than doing
192 it just once per client session. This is useful to measure the
193 connection overhead.
194 -d
196 --debug
197 Print debugging output.
198 -D varname=value
200 --define=varname=value
201 Define a variable for use by a custom script (see below). Multiple
202 -D options are allowed.
203 -f filename[@weight]
205 --file=filename[@weight]
206 Add a transaction script read from filename to the list of scripts
207 to be executed.
208 Optionally, write an integer weight after @ to adjust the
210 probability of selecting this script versus other ones. The default
211 weight is 1. (To use a script file name that includes an @
212 character, append a weight so that there is no ambiguity, for
213 example filen@me@1.) See below for details.
214 -j threads
216 --jobs=threads
217 Number of worker threads within pgbench. Using more than one thread
218 can be helpful on multi-CPU machines. Clients are distributed as
219 evenly as possible among available threads. Default is 1.
220 -l
222 --log
223 Write information about each transaction to a log file. See below
224 for details.
225 -L limit
227 --latency-limit=limit
228 Transactions that last more than limit milliseconds are counted and
229 reported separately, as late.
230 When throttling is used (--rate=...), transactions that lag behind
232 schedule by more than limit ms, and thus have no hope of meeting
233 the latency limit, are not sent to the server at all. They are
234 counted and reported separately as skipped.
235 -M querymode
237 --protocol=querymode
238 Protocol to use for submitting queries to the server:
239 · simple: use simple query protocol.
241 · extended: use extended query protocol.
243 · prepared: use extended query protocol with prepared statements.
245 In the prepared mode, pgbench reuses the parse analysis result
247 starting from the second query iteration, so pgbench runs faster
248 than in other modes.
249 The default is simple query protocol. (See Chapter 52 for more
251 information.)
252 -n
254 --no-vacuum
255 Perform no vacuuming before running the test. This option is
256 necessary if you are running a custom test scenario that does not
257 include the standard tables pgbench_accounts, pgbench_branches,
258 pgbench_history, and pgbench_tellers.
259 -N
261 --skip-some-updates
262 Run built-in simple-update script. Shorthand for -b simple-update.
263 -P sec
265 --progress=sec
266 Show progress report every sec seconds. The report includes the
267 time since the beginning of the run, the TPS since the last report,
268 and the transaction latency average and standard deviation since
269 the last report. Under throttling (-R), the latency is computed
270 with respect to the transaction scheduled start time, not the
271 actual transaction beginning time, thus it also includes the
272 average schedule lag time.
273 -r
275 --report-latencies
276 Report the average per-statement latency (execution time from the
277 perspective of the client) of each command after the benchmark
278 finishes. See below for details.
279 -R rate
281 --rate=rate
282 Execute transactions targeting the specified rate instead of
283 running as fast as possible (the default). The rate is given in
284 transactions per second. If the targeted rate is above the maximum
285 possible rate, the rate limit won't impact the results.
286 The rate is targeted by starting transactions along a
288 Poisson-distributed schedule time line. The expected start time
289 schedule moves forward based on when the client first started, not
290 when the previous transaction ended. That approach means that when
291 transactions go past their original scheduled end time, it is
292 possible for later ones to catch up again.
293 When throttling is active, the transaction latency reported at the
295 end of the run is calculated from the scheduled start times, so it
296 includes the time each transaction had to wait for the previous
297 transaction to finish. The wait time is called the schedule lag
298 time, and its average and maximum are also reported separately. The
299 transaction latency with respect to the actual transaction start
300 time, i.e., the time spent executing the transaction in the
301 database, can be computed by subtracting the schedule lag time from
302 the reported latency.
303 If --latency-limit is used together with --rate, a transaction can
305 lag behind so much that it is already over the latency limit when
306 the previous transaction ends, because the latency is calculated
307 from the scheduled start time. Such transactions are not sent to
308 the server, but are skipped altogether and counted separately.
309 A high schedule lag time is an indication that the system cannot
311 process transactions at the specified rate, with the chosen number
312 of clients and threads. When the average transaction execution time
313 is longer than the scheduled interval between each transaction,
314 each successive transaction will fall further behind, and the
315 schedule lag time will keep increasing the longer the test run is.
316 When that happens, you will have to reduce the specified
317 transaction rate.
318 -s scale_factor
320 --scale=scale_factor
321 Report the specified scale factor in pgbench's output. With the
322 built-in tests, this is not necessary; the correct scale factor
323 will be detected by counting the number of rows in the
324 pgbench_branches table. However, when testing only custom
325 benchmarks (-f option), the scale factor will be reported as 1
326 unless this option is used.
327 -S
329 --select-only
330 Run built-in select-only script. Shorthand for -b select-only.
331 -t transactions
333 --transactions=transactions
334 Number of transactions each client runs. Default is 10.
335 -T seconds
337 --time=seconds
338 Run the test for this many seconds, rather than a fixed number of
339 transactions per client. -t and -T are mutually exclusive.
340 -v
342 --vacuum-all
343 Vacuum all four standard tables before running the test. With
344 neither -n nor -v, pgbench will vacuum the pgbench_tellers and
345 pgbench_branches tables, and will truncate pgbench_history.
346 --aggregate-interval=seconds
348 Length of aggregation interval (in seconds). May be used only with
349 -l option. With this option, the log contains per-interval summary
350 data, as described below.
351 --log-prefix=prefix
353 Set the filename prefix for the log files created by --log. The
354 default is pgbench_log.
355 --progress-timestamp
357 When showing progress (option -P), use a timestamp (Unix epoch)
358 instead of the number of seconds since the beginning of the run.
359 The unit is in seconds, with millisecond precision after the dot.
360 This helps compare logs generated by various tools.
361 --random-seed=SEED
363 Set random generator seed. Seeds the system random number
364 generator, which then produces a sequence of initial generator
365 states, one for each thread. Values for SEED may be: time (the
366 default, the seed is based on the current time), rand (use a strong
367 random source, failing if none is available), or an unsigned
368 decimal integer value. The random generator is invoked explicitly
369 from a pgbench script (random... functions) or implicitly (for
370 instance option --rate uses it to schedule transactions). When
371 explicitly set, the value used for seeding is shown on the
372 terminal. Any value allowed for SEED may also be provided through
373 the environment variable PGBENCH_RANDOM_SEED. To ensure that the
374 provided seed impacts all possible uses, put this option first or
375 use the environment variable.
376 Setting the seed explicitly allows to reproduce a pgbench run
378 exactly, as far as random numbers are concerned. As the random
379 state is managed per thread, this means the exact same pgbench run
380 for an identical invocation if there is one client per thread and
381 there are no external or data dependencies. From a statistical
382 viewpoint reproducing runs exactly is a bad idea because it can
383 hide the performance variability or improve performance unduly,
384 e.g., by hitting the same pages as a previous run. However, it may
385 also be of great help for debugging, for instance re-running a
386 tricky case which leads to an error. Use wisely.
387 --sampling-rate=rate
389 Sampling rate, used when writing data into the log, to reduce the
390 amount of log generated. If this option is given, only the
391 specified fraction of transactions are logged. 1.0 means all
392 transactions will be logged, 0.05 means only 5% of the transactions
393 will be logged.
394 Remember to take the sampling rate into account when processing the
396 log file. For example, when computing TPS values, you need to
397 multiply the numbers accordingly (e.g., with 0.01 sample rate,
398 you'll only get 1/100 of the actual TPS).
399 Common Options
401 pgbench accepts the following command-line common arguments:
402 -h hostname
404 --host=hostname
405 The database server's host name
406 -p port
408 --port=port
409 The database server's port number
410 -U login
412 --username=login
413 The user name to connect as
414 -V
416 --version
417 Print the pgbench version and exit.
418 -?
420 --help
421 Show help about pgbench command line arguments, and exit.
422
424 A successful run will exit with status 0. Exit status 1 indicates
425 static problems such as invalid command-line options. Errors during the
426 run such as database errors or problems in the script will result in
427 exit status 2. In the latter case, pgbench will print partial results.
428
430 PGHOST
431 PGPORT
432 PGUSER
433 Default connection parameters.
434 This utility, like most other PostgreSQL utilities, uses the
436 environment variables supported by libpq (see Section 33.14).
437
439 What Is the “Transaction” Actually Performed in pgbench?
440 pgbench executes test scripts chosen randomly from a specified list.
441 The scripts may include built-in scripts specified with -b and
442 user-provided scripts specified with -f. Each script may be given a
443 relative weight specified after an @ so as to change its selection
444 probability. The default weight is 1. Scripts with a weight of 0 are
445 ignored.
446 The default built-in transaction script (also invoked with -b
448 tpcb-like) issues seven commands per transaction over randomly chosen
449 aid, tid, bid and delta. The scenario is inspired by the TPC-B
450 benchmark, but is not actually TPC-B, hence the name.
451 1. BEGIN;
453 2. UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid
455 = :aid;
456 3. SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
458 4. UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid =
460 :tid;
461 5. UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid
463 = :bid;
464 6. INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES
466 (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
467 7. END;
469 If you select the simple-update built-in (also -N), steps 4 and 5
471 aren't included in the transaction. This will avoid update contention
472 on these tables, but it makes the test case even less like TPC-B.
473 If you select the select-only built-in (also -S), only the SELECT is
475 issued.
476 Custom Scripts
478 pgbench has support for running custom benchmark scenarios by replacing
479 the default transaction script (described above) with a transaction
480 script read from a file (-f option). In this case a “transaction”
481 counts as one execution of a script file.
482 A script file contains one or more SQL commands terminated by
484 semicolons. Empty lines and lines beginning with -- are ignored. Script
485 files can also contain “meta commands”, which are interpreted by
486 pgbench itself, as described below.
487 Note
489 Before PostgreSQL 9.6, SQL commands in script files were terminated
490 by newlines, and so they could not be continued across lines. Now a
491 semicolon is required to separate consecutive SQL commands (though
492 a SQL command does not need one if it is followed by a meta
493 command). If you need to create a script file that works with both
494 old and new versions of pgbench, be sure to write each SQL command
495 on a single line ending with a semicolon.
496 There is a simple variable-substitution facility for script files.
498 Variable names must consist of letters (including non-Latin letters),
499 digits, and underscores, with the first character not being a digit.
500 Variables can be set by the command-line -D option, explained above, or
501 by the meta commands explained below. In addition to any variables
502 preset by -D command-line options, there are a few variables that are
503 preset automatically, listed in Table 258. A value specified for these
504 variables using -D takes precedence over the automatic presets. Once
505 set, a variable's value can be inserted into a SQL command by writing
506 :variablename. When running more than one client session, each session
507 has its own set of variables. pgbench supports up to 255 variable uses
508 in one statement.
509 Table 258. Automatic Variables
511 ┌─────────────┬────────────────────────────┐
512 │Variable │ Description │
513 ├─────────────┼────────────────────────────┤
514 │client_id │ unique number identifying │
515 │ │ the client session (starts │
516 │ │ from zero) │
517 ├─────────────┼────────────────────────────┤
518 │default_seed │ seed used in hash │
519 │ │ functions by default │
520 ├─────────────┼────────────────────────────┤
521 │random_seed │ random generator seed │
522 │ │ (unless overwritten with │
523 │ │ -D) │
524 ├─────────────┼────────────────────────────┤
525 │scale │ current scale factor │
526 └─────────────┴────────────────────────────┘
527 Script file meta commands begin with a backslash (\) and normally
529 extend to the end of the line, although they can be continued to
530 additional lines by writing backslash-return. Arguments to a meta
531 command are separated by white space. These meta commands are
532 supported:
533 \gset [prefix]
535 This command may be used to end SQL queries, taking the place of
536 the terminating semicolon (;).
537 When this command is used, the preceding SQL query is expected to
539 return one row, the columns of which are stored into variables
540 named after column names, and prefixed with prefix if provided.
541 The following example puts the final account balance from the first
543 query into variable abalance, and fills variables p_two and p_three
544 with integers from the third query. The result of the second query
545 is discarded.
546 UPDATE pgbench_accounts
548 SET abalance = abalance + :delta
549 WHERE aid = :aid
550 RETURNING abalance \gset
551 -- compound of two queries
552 SELECT 1 \;
553 SELECT 2 AS two, 3 AS three \gset p_
554 \if expression
557 \elif expression
558 \else
559 \endif
560 This group of commands implements nestable conditional blocks,
561 similarly to psql's \if expression. Conditional expressions are
562 identical to those with \set, with non-zero values interpreted as
563 true.
564 \set varname expression
566 Sets variable varname to a value calculated from expression. The
567 expression may contain the NULL constant, Boolean constants TRUE
568 and FALSE, integer constants such as 5432, double constants such as
569 3.14159, references to variables :variablename, operators with
570 their usual SQL precedence and associativity, function calls, SQL
571 CASE generic conditional expressions and parentheses.
572 Functions and most operators return NULL on NULL input.
574 For conditional purposes, non zero numerical values are TRUE, zero
576 numerical values and NULL are FALSE.
577 Too large or small integer and double constants, as well as integer
579 arithmetic operators (+, -, * and /) raise errors on overflows.
580 When no final ELSE clause is provided to a CASE, the default value
582 is NULL.
583 Examples:
585 \set ntellers 10 * :scale
587 \set aid (1021 * random(1, 100000 * :scale)) % \
588 (100000 * :scale) + 1
589 \set divx CASE WHEN :x <> 0 THEN :y/:x ELSE NULL END
590 \sleep number [ us | ms | s ]
592 Causes script execution to sleep for the specified duration in
593 microseconds (us), milliseconds (ms) or seconds (s). If the unit is
594 omitted then seconds are the default. number can be either an
595 integer constant or a :variablename reference to a variable having
596 an integer value.
597 Example:
599 \sleep 10 ms
601 \setshell varname command [ argument ... ]
603 Sets variable varname to the result of the shell command command
604 with the given argument(s). The command must return an integer
605 value through its standard output.
606 command and each argument can be either a text constant or a
608 :variablename reference to a variable. If you want to use an
609 argument starting with a colon, write an additional colon at the
610 beginning of argument.
611 Example:
613 \setshell variable_to_be_assigned command literal_argument :variable ::literal_starting_with_colon
615 \shell command [ argument ... ]
617 Same as \setshell, but the result of the command is discarded.
618 Example:
620 \shell command literal_argument :variable ::literal_starting_with_colon
622 Built-in Operators
624 The arithmetic, bitwise, comparison and logical operators listed in
625 Table 259 are built into pgbench and may be used in expressions
626 appearing in \set.
627 Table 259. pgbench Operators by Increasing Precedence
629 ┌──────────────────┬──────────────────┬───────────┬────────┐
630 │Operator │ Description │ Example │ Result │
631 ├──────────────────┼──────────────────┼───────────┼────────┤
632 │OR │ logical or │ 5 or 0 │ TRUE │
633 ├──────────────────┼──────────────────┼───────────┼────────┤
634 │AND │ logical and │ 3 and 0 │ FALSE │
635 ├──────────────────┼──────────────────┼───────────┼────────┤
636 │NOT │ logical not │ not false │ TRUE │
637 ├──────────────────┼──────────────────┼───────────┼────────┤
638 │IS [NOT] │ value tests │ 1 is null │ FALSE │
639 │(NULL|TRUE|FALSE) │ │ │ │
640 ├──────────────────┼──────────────────┼───────────┼────────┤
641 │ISNULL|NOTNULL │ null tests │ 1 notnull │ TRUE │
642 ├──────────────────┼──────────────────┼───────────┼────────┤
643 │= │ is equal │ 5 = 4 │ FALSE │
644 ├──────────────────┼──────────────────┼───────────┼────────┤
645 │<> │ is not equal │ 5 <> 4 │ TRUE │
646 ├──────────────────┼──────────────────┼───────────┼────────┤
647 │!= │ is not equal │ 5 != 5 │ FALSE │
648 ├──────────────────┼──────────────────┼───────────┼────────┤
649 │< │ lower than │ 5 < 4 │ FALSE │
650 ├──────────────────┼──────────────────┼───────────┼────────┤
651 │<= │ lower or equal │ 5 <= 4 │ FALSE │
652 ├──────────────────┼──────────────────┼───────────┼────────┤
653 │> │ greater than │ 5 > 4 │ TRUE │
654 ├──────────────────┼──────────────────┼───────────┼────────┤
655 │>= │ greater or equal │ 5 >= 4 │ TRUE │
656 ├──────────────────┼──────────────────┼───────────┼────────┤
657 │| │ integer bitwise │ 1 | 2 │ 3 │
658 │ │ OR │ │ │
659 ├──────────────────┼──────────────────┼───────────┼────────┤
660 │# │ integer bitwise │ 1 # 3 │ 2 │
661 │ │ XOR │ │ │
662 ├──────────────────┼──────────────────┼───────────┼────────┤
663 │& │ integer bitwise │ 1 & 3 │ 1 │
664 │ │ AND │ │ │
665 ├──────────────────┼──────────────────┼───────────┼────────┤
666 │~ │ integer bitwise │ ~ 1 │ -2 │
667 │ │ NOT │ │ │
668 ├──────────────────┼──────────────────┼───────────┼────────┤
669 │<< │ integer bitwise │ 1 << 2 │ 4 │
670 │ │ shift left │ │ │
671 ├──────────────────┼──────────────────┼───────────┼────────┤
672 │>> │ integer bitwise │ 8 >> 2 │ 2 │
673 │ │ shift right │ │ │
674 ├──────────────────┼──────────────────┼───────────┼────────┤
675 │+ │ addition │ 5 + 4 │ 9 │
676 ├──────────────────┼──────────────────┼───────────┼────────┤
677 │- │ subtraction │ 3 - 2.0 │ 1.0 │
678 ├──────────────────┼──────────────────┼───────────┼────────┤
679 │* │ multiplication │ 5 * 4 │ 20 │
680 ├──────────────────┼──────────────────┼───────────┼────────┤
681 │/ │ division │ 5 / 3 │ 1 │
682 │ │ (integer │ │ │
683 │ │ truncates the │ │ │
684 │ │ results) │ │ │
685 ├──────────────────┼──────────────────┼───────────┼────────┤
686 │% │ modulo │ 3 % 2 │ 1 │
687 ├──────────────────┼──────────────────┼───────────┼────────┤
688 │- │ opposite │ - 2.0 │ -2.0 │
689 └──────────────────┴──────────────────┴───────────┴────────┘
690 Built-In Functions
692 The functions listed in Table 260 are built into pgbench and may be
693 used in expressions appearing in \set.
694 Table 260. pgbench Functions
696 ┌───────────────────────┬───────────────┬───────────────────────────┬───────────────────────┬────────────────────────┐
697 │Function │ Return Type │ Description │ Example │ Result │
698 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
699 │abs(a) │ same as a │ absolute │ abs(-17) │ 17 │
700 │ │ │ value │ │ │
701 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
702 │debug(a) │ same as a │ print a to │ debug(5432.1) │ 5432.1 │
703 │ │ │ stderr, │ │ │
704 │ │ │ and │ │ │
705 │ │ │ return a │ │ │
706 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
707 │double(i) │ double │ cast to │ double(5432) │ 5432.0 │
708 │ │ │ double │ │ │
709 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
710 │exp(x) │ double │ exponential │ exp(1.0) │ 2.718281828459045 │
711 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
712 │greatest(a [, │ double if any │ largest value │ greatest(5, │ 5 │
713 │... ] ) │ a is double, │ among │ 4, 3, 2) │ │
714 │ │ else integer │ arguments │ │ │
715 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
716 │hash(a [, │ integer │ alias for │ hash(10, │ -5817877081768721676 │
717 │seed ] ) │ │ hash_murmur2() │ 5432) │ │
718 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
719 │hash_fnv1a(a │ integer │ FNV-1a hash │ hash_fnv1a(10, │ -7793829335365542153 │
720 │[, seed ] ) │ │ │ 5432) │ │
721 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
722 │hash_murmur2(a │ integer │ MurmurHash2 │ hash_murmur2(10, │ -5817877081768721676 │
723 │[, seed ] ) │ │ hash │ 5432) │ │
724 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
725 │int(x) │ integer │ cast to int │ int(5.4 + 3.8) │ 9 │
726 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
727 │least(a [, ... │ double if any │ smallest value │ least(5, 4, 3, │ 2.1 │
728 │] ) │ a is double, │ among │ 2.1) │ │
729 │ │ else integer │ arguments │ │ │
730 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
731 │ln(x) │ double │ natural │ ln(2.718281828459045) │ 1.0 │
732 │ │ │ logarithm │ │ │
733 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
734 │mod(i, j) │ integer │ modulo │ mod(54, 32) │ 22 │
735 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
736 │pi() │ double │ value of the │ pi() │ 3.14159265358979323846 │
737 │ │ │ constant PI │ │ │
738 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
739 │pow(x, y), │ double │ exponentiation │ pow(2.0, 10), │ 1024.0 │
740 │power(x, y) │ │ │ power(2.0, 10) │ │
741 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
742 │random(lb, ub) │ integer │ uniformly-distributed │ random(1, 10) │ an integer between 1 │
743 │ │ │ random integer │ │ and 10 │
744 │ │ │ in [lb, ub] │ │ │
745 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
746 │random_exponential(lb, │ integer │ exponentially-distributed │ random_exponential(1, │ an integer between 1 │
747 │ub, parameter) │ │ random integer in │ 10, 3.0) │ and 10 │
748 │ │ │ [lb, ub], │ │ │
749 │ │ │ see │ │ │
750 │ │ │ below │ │ │
751 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
752 │random_gaussian(lb, │ integer │ Gaussian-distributed │ random_gaussian(1, │ an integer between 1 │
753 │ub, parameter) │ │ random integer in [lb, │ 10, 2.5) │ and 10 │
754 │ │ │ ub], │ │ │
755 │ │ │ see below │ │ │
756 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
757 │random_zipfian(lb, ub, │ integer │ Zipfian-distributed │ random_zipfian(1, 10, │ an integer between 1 │
758 │parameter) │ │ random integer in [lb, │ 1.5) │ and 10 │
759 │ │ │ ub], │ │ │
760 │ │ │ see below │ │ │
761 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
762 │sqrt(x) │ double │ square root │ sqrt(2.0) │ 1.414213562 │
763 └───────────────────────┴───────────────┴───────────────────────────┴───────────────────────┴────────────────────────┘
764 The random function generates values using a uniform distribution, that
766 is all the values are drawn within the specified range with equal
767 probability. The random_exponential, random_gaussian and random_zipfian
768 functions require an additional double parameter which determines the
769 precise shape of the distribution.
770 · For an exponential distribution, parameter controls the
772 distribution by truncating a quickly-decreasing exponential
773 distribution at parameter, and then projecting onto integers
774 between the bounds. To be precise, with
775 f(x) = exp(-parameter * (x - min) / (max - min + 1)) / (1 - exp(-parameter))
777 Then value i between min and max inclusive is drawn with
779 probability: f(i) - f(i + 1).
780 Intuitively, the larger the parameter, the more frequently values
782 close to min are accessed, and the less frequently values close to
783 max are accessed. The closer to 0 parameter is, the flatter (more
784 uniform) the access distribution. A crude approximation of the
785 distribution is that the most frequent 1% values in the range,
786 close to min, are drawn parameter% of the time. The parameter value
787 must be strictly positive.
788 · For a Gaussian distribution, the interval is mapped onto a standard
790 normal distribution (the classical bell-shaped Gaussian curve)
791 truncated at -parameter on the left and +parameter on the right.
792 Values in the middle of the interval are more likely to be drawn.
793 To be precise, if PHI(x) is the cumulative distribution function of
794 the standard normal distribution, with mean mu defined as (max +
795 min) / 2.0, with
796 f(x) = PHI(2.0 * parameter * (x - mu) / (max - min + 1)) /
798 (2.0 * PHI(parameter) - 1)
799 then value i between min and max inclusive is drawn with
801 probability: f(i + 0.5) - f(i - 0.5). Intuitively, the larger the
802 parameter, the more frequently values close to the middle of the
803 interval are drawn, and the less frequently values close to the min
804 and max bounds. About 67% of values are drawn from the middle 1.0 /
805 parameter, that is a relative 0.5 / parameter around the mean, and
806 95% in the middle 2.0 / parameter, that is a relative 1.0 /
807 parameter around the mean; for instance, if parameter is 4.0, 67%
808 of values are drawn from the middle quarter (1.0 / 4.0) of the
809 interval (i.e., from 3.0 / 8.0 to 5.0 / 8.0) and 95% from the
810 middle half (2.0 / 4.0) of the interval (second and third
811 quartiles). The minimum allowed parameter value is 2.0.
812 · random_zipfian generates a bounded Zipfian distribution. parameter
814 defines how skewed the distribution is. The larger the parameter,
815 the more frequently values closer to the beginning of the interval
816 are drawn. The distribution is such that, assuming the range starts
817 from 1, the ratio of the probability of drawing k versus drawing
818 k+1 is ((k+1)/k)**parameter. For example, random_zipfian(1, ...,
819 2.5) produces the value 1 about (2/1)**2.5 = 5.66 times more
820 frequently than 2, which itself is produced (3/2)**2.5 = 2.76 times
821 more frequently than 3, and so on.
822 pgbench's implementation is based on "Non-Uniform Random Variate
824 Generation", Luc Devroye, p. 550-551, Springer 1986. Due to
825 limitations of that algorithm, the parameter value is restricted to
826 the range [1.001, 1000].
827 Hash functions hash, hash_murmur2 and hash_fnv1a accept an input value
829 and an optional seed parameter. In case the seed isn't provided the
830 value of :default_seed is used, which is initialized randomly unless
831 set by the command-line -D option. Hash functions can be used to
832 scatter the distribution of random functions such as random_zipfian or
833 random_exponential. For instance, the following pgbench script
834 simulates possible real world workload typical for social media and
835 blogging platforms where few accounts generate excessive load:
836 \set r random_zipfian(0, 100000000, 1.07)
838 \set k abs(hash(:r)) % 1000000
839 In some cases several distinct distributions are needed which don't
841 correlate with each other and this is when implicit seed parameter
842 comes in handy:
843 \set k1 abs(hash(:r, :default_seed + 123)) % 1000000
845 \set k2 abs(hash(:r, :default_seed + 321)) % 1000000
846 As an example, the full definition of the built-in TPC-B-like
848 transaction is:
849 \set aid random(1, 100000 * :scale)
851 \set bid random(1, 1 * :scale)
852 \set tid random(1, 10 * :scale)
853 \set delta random(-5000, 5000)
854 BEGIN;
855 UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
856 SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
857 UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
858 UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
859 INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
860 END;
861 This script allows each iteration of the transaction to reference
863 different, randomly-chosen rows. (This example also shows why it's
864 important for each client session to have its own variables — otherwise
865 they'd not be independently touching different rows.)
866 Per-Transaction Logging
868 With the -l option (but without the --aggregate-interval option),
869 pgbench writes information about each transaction to a log file. The
870 log file will be named prefix.nnn, where prefix defaults to
871 pgbench_log, and nnn is the PID of the pgbench process. The prefix can
872 be changed by using the --log-prefix option. If the -j option is 2 or
873 higher, so that there are multiple worker threads, each will have its
874 own log file. The first worker will use the same name for its log file
875 as in the standard single worker case. The additional log files for the
876 other workers will be named prefix.nnn.mmm, where mmm is a sequential
877 number for each worker starting with 1.
878 The format of the log is:
880 client_id transaction_no time script_no time_epoch time_us [ schedule_lag ]
882 where client_id indicates which client session ran the transaction,
884 transaction_no counts how many transactions have been run by that
885 session, time is the total elapsed transaction time in microseconds,
886 script_no identifies which script file was used (useful when multiple
887 scripts were specified with -f or -b), and time_epoch/time_us are a
888 Unix-epoch time stamp and an offset in microseconds (suitable for
889 creating an ISO 8601 time stamp with fractional seconds) showing when
890 the transaction completed. The schedule_lag field is the difference
891 between the transaction's scheduled start time, and the time it
892 actually started, in microseconds. It is only present when the --rate
893 option is used. When both --rate and --latency-limit are used, the time
894 for a skipped transaction will be reported as skipped.
895 Here is a snippet of a log file generated in a single-client run:
897 0 199 2241 0 1175850568 995598
899 0 200 2465 0 1175850568 998079
900 0 201 2513 0 1175850569 608
901 0 202 2038 0 1175850569 2663
902 Another example with --rate=100 and --latency-limit=5 (note the
904 additional schedule_lag column):
905 0 81 4621 0 1412881037 912698 3005
907 0 82 6173 0 1412881037 914578 4304
908 0 83 skipped 0 1412881037 914578 5217
909 0 83 skipped 0 1412881037 914578 5099
910 0 83 4722 0 1412881037 916203 3108
911 0 84 4142 0 1412881037 918023 2333
912 0 85 2465 0 1412881037 919759 740
913 In this example, transaction 82 was late, because its latency (6.173
915 ms) was over the 5 ms limit. The next two transactions were skipped,
916 because they were already late before they were even started.
917 When running a long test on hardware that can handle a lot of
919 transactions, the log files can become very large. The --sampling-rate
920 option can be used to log only a random sample of transactions.
921 Aggregated Logging
923 With the --aggregate-interval option, a different format is used for
924 the log files:
925 interval_start num_transactions sum_latency sum_latency_2 min_latency max_latency [ sum_lag sum_lag_2 min_lag max_lag [ skipped ] ]
927 where interval_start is the start of the interval (as a Unix epoch time
929 stamp), num_transactions is the number of transactions within the
930 interval, sum_latency is the sum of the transaction latencies within
931 the interval, sum_latency_2 is the sum of squares of the transaction
932 latencies within the interval, min_latency is the minimum latency
933 within the interval, and max_latency is the maximum latency within the
934 interval. The next fields, sum_lag, sum_lag_2, min_lag, and max_lag,
935 are only present if the --rate option is used. They provide statistics
936 about the time each transaction had to wait for the previous one to
937 finish, i.e., the difference between each transaction's scheduled start
938 time and the time it actually started. The very last field, skipped, is
939 only present if the --latency-limit option is used, too. It counts the
940 number of transactions skipped because they would have started too
941 late. Each transaction is counted in the interval when it was
942 committed.
943 Here is some example output:
945 1345828501 5601 1542744 483552416 61 2573
947 1345828503 7884 1979812 565806736 60 1479
948 1345828505 7208 1979422 567277552 59 1391
949 1345828507 7685 1980268 569784714 60 1398
950 1345828509 7073 1979779 573489941 236 1411
951 Notice that while the plain (unaggregated) log file shows which script
953 was used for each transaction, the aggregated log does not. Therefore
954 if you need per-script data, you need to aggregate the data on your
955 own.
956 Per-Statement Latencies
958 With the -r option, pgbench collects the elapsed transaction time of
959 each statement executed by every client. It then reports an average of
960 those values, referred to as the latency for each statement, after the
961 benchmark has finished.
962 For the default script, the output will look similar to this:
964 starting vacuum...end.
966 transaction type: <builtin: TPC-B (sort of)>
967 scaling factor: 1
968 query mode: simple
969 number of clients: 10
970 number of threads: 1
971 number of transactions per client: 1000
972 number of transactions actually processed: 10000/10000
973 latency average = 15.844 ms
974 latency stddev = 2.715 ms
975 tps = 618.764555 (including connections establishing)
976 tps = 622.977698 (excluding connections establishing)
977 statement latencies in milliseconds:
978 0.002 \set aid random(1, 100000 * :scale)
979 0.005 \set bid random(1, 1 * :scale)
980 0.002 \set tid random(1, 10 * :scale)
981 0.001 \set delta random(-5000, 5000)
982 0.326 BEGIN;
983 0.603 UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
984 0.454 SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
985 5.528 UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
986 7.335 UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
987 0.371 INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
988 1.212 END;
989 If multiple script files are specified, the averages are reported
991 separately for each script file.
992 Note that collecting the additional timing information needed for
994 per-statement latency computation adds some overhead. This will slow
995 average execution speed and lower the computed TPS. The amount of
996 slowdown varies significantly depending on platform and hardware.
997 Comparing average TPS values with and without latency reporting enabled
998 is a good way to measure if the timing overhead is significant.
999 Good Practices
1001 It is very easy to use pgbench to produce completely meaningless
1002 numbers. Here are some guidelines to help you get useful results.
1003 In the first place, never believe any test that runs for only a few
1005 seconds. Use the -t or -T option to make the run last at least a few
1006 minutes, so as to average out noise. In some cases you could need hours
1007 to get numbers that are reproducible. It's a good idea to try the test
1008 run a few times, to find out if your numbers are reproducible or not.
1009 For the default TPC-B-like test scenario, the initialization scale
1011 factor (-s) should be at least as large as the largest number of
1012 clients you intend to test (-c); else you'll mostly be measuring update
1013 contention. There are only -s rows in the pgbench_branches table, and
1014 every transaction wants to update one of them, so -c values in excess
1015 of -s will undoubtedly result in lots of transactions blocked waiting
1016 for other transactions.
1017 The default test scenario is also quite sensitive to how long it's been
1019 since the tables were initialized: accumulation of dead rows and dead
1020 space in the tables changes the results. To understand the results you
1021 must keep track of the total number of updates and when vacuuming
1022 happens. If autovacuum is enabled it can result in unpredictable
1023 changes in measured performance.
1024 A limitation of pgbench is that it can itself become the bottleneck
1026 when trying to test a large number of client sessions. This can be
1027 alleviated by running pgbench on a different machine from the database
1028 server, although low network latency will be essential. It might even
1029 be useful to run several pgbench instances concurrently, on several
1030 client machines, against the same database server.
1031 Security
1033 If untrusted users have access to a database that has not adopted a
1034 secure schema usage pattern, do not run pgbench in that database.
1035 pgbench uses unqualified names and does not manipulate the search path.
1036PostgreSQL 12.6 2021 PGBENCH(1)