1PGBENCH(1) PostgreSQL 12.2 Documentation PGBENCH(1)
2
3
4
6 pgbench - run a benchmark test on PostgreSQL
7
9 pgbench -i [option...] [dbname]
10
11 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
22 Typical output from pgbench looks like:
23
24 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
34 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
43 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
49 pgbench -i [ other-options ] dbname
50
51 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
55 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
61 At the default “scale factor” of 1, the tables initially contain this
62 many rows:
63
64 table # of rows
65 ---------------------------------
66 pgbench_branches 1
67 pgbench_tellers 10
68 pgbench_accounts 100000
69 pgbench_history 0
70
71 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
75 Once you have done the necessary setup, you can run your benchmark with
76 a command that doesn't include -i, that is
77
78 pgbench [ options ] dbname
79
80 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
90 Initialization Options
91 pgbench accepts the following command-line initialization arguments:
92
93 -i
94 --initialize
95 Required to invoke initialization mode.
96
97 -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
104 d (Drop)
105 Drop any existing pgbench tables.
106
107 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
112 g (Generate data)
113 Generate data and load it into the standard tables, replacing
114 any data already present.
115
116 v (Vacuum)
117 Invoke VACUUM on the standard tables.
118
119 p (create Primary keys)
120 Create primary key indexes on the standard tables.
121
122 f (create Foreign keys)
123 Create foreign key constraints between the standard tables.
124 (Note that this step is not performed by default.)
125
126
127 -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
132 -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
137 -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
144 -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
153 --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
158 --index-tablespace=index_tablespace
159 Create indexes in the specified tablespace, rather than the default
160 tablespace.
161
162 --tablespace=tablespace
163 Create tables in the specified tablespace, rather than the default
164 tablespace.
165
166 --unlogged-tables
167 Create all tables as unlogged tables, rather than permanent tables.
168
169 Benchmarking Options
170 pgbench accepts the following command-line benchmarking arguments:
171
172 -b scriptname[@weight]
173 --builtin=scriptname[@weight]
174 Add the specified built-in script to the list of executed scripts.
175 An optional integer weight after @ allows to adjust the probability
176 of drawing the script. If not specified, it is set to 1. Available
177 built-in scripts are: tpcb-like, simple-update and select-only.
178 Unambiguous prefixes of built-in names are accepted. With special
179 name list, show the list of built-in scripts and exit immediately.
180
181 -c clients
182 --client=clients
183 Number of clients simulated, that is, number of concurrent database
184 sessions. Default is 1.
185
186 -C
187 --connect
188 Establish a new connection for each transaction, rather than doing
189 it just once per client session. This is useful to measure the
190 connection overhead.
191
192 -d
193 --debug
194 Print debugging output.
195
196 -D varname=value
197 --define=varname=value
198 Define a variable for use by a custom script (see below). Multiple
199 -D options are allowed.
200
201 -f filename[@weight]
202 --file=filename[@weight]
203 Add a transaction script read from filename to the list of executed
204 scripts. An optional integer weight after @ allows to adjust the
205 probability of drawing the test. See below for details.
206
207 -j threads
208 --jobs=threads
209 Number of worker threads within pgbench. Using more than one thread
210 can be helpful on multi-CPU machines. Clients are distributed as
211 evenly as possible among available threads. Default is 1.
212
213 -l
214 --log
215 Write information about each transaction to a log file. See below
216 for details.
217
218 -L limit
219 --latency-limit=limit
220 Transactions that last more than limit milliseconds are counted and
221 reported separately, as late.
222
223 When throttling is used (--rate=...), transactions that lag behind
224 schedule by more than limit ms, and thus have no hope of meeting
225 the latency limit, are not sent to the server at all. They are
226 counted and reported separately as skipped.
227
228 -M querymode
229 --protocol=querymode
230 Protocol to use for submitting queries to the server:
231
232 · simple: use simple query protocol.
233
234 · extended: use extended query protocol.
235
236 · prepared: use extended query protocol with prepared statements.
237
238 In the prepared mode, pgbench reuses the parse analysis result
239 starting from the second query iteration, so pgbench runs faster
240 than in other modes.
241
242 The default is simple query protocol. (See Chapter 52 for more
243 information.)
244
245 -n
246 --no-vacuum
247 Perform no vacuuming before running the test. This option is
248 necessary if you are running a custom test scenario that does not
249 include the standard tables pgbench_accounts, pgbench_branches,
250 pgbench_history, and pgbench_tellers.
251
252 -N
253 --skip-some-updates
254 Run built-in simple-update script. Shorthand for -b simple-update.
255
256 -P sec
257 --progress=sec
258 Show progress report every sec seconds. The report includes the
259 time since the beginning of the run, the TPS since the last report,
260 and the transaction latency average and standard deviation since
261 the last report. Under throttling (-R), the latency is computed
262 with respect to the transaction scheduled start time, not the
263 actual transaction beginning time, thus it also includes the
264 average schedule lag time.
265
266 -r
267 --report-latencies
268 Report the average per-statement latency (execution time from the
269 perspective of the client) of each command after the benchmark
270 finishes. See below for details.
271
272 -R rate
273 --rate=rate
274 Execute transactions targeting the specified rate instead of
275 running as fast as possible (the default). The rate is given in
276 transactions per second. If the targeted rate is above the maximum
277 possible rate, the rate limit won't impact the results.
278
279 The rate is targeted by starting transactions along a
280 Poisson-distributed schedule time line. The expected start time
281 schedule moves forward based on when the client first started, not
282 when the previous transaction ended. That approach means that when
283 transactions go past their original scheduled end time, it is
284 possible for later ones to catch up again.
285
286 When throttling is active, the transaction latency reported at the
287 end of the run is calculated from the scheduled start times, so it
288 includes the time each transaction had to wait for the previous
289 transaction to finish. The wait time is called the schedule lag
290 time, and its average and maximum are also reported separately. The
291 transaction latency with respect to the actual transaction start
292 time, i.e. the time spent executing the transaction in the
293 database, can be computed by subtracting the schedule lag time from
294 the reported latency.
295
296 If --latency-limit is used together with --rate, a transaction can
297 lag behind so much that it is already over the latency limit when
298 the previous transaction ends, because the latency is calculated
299 from the scheduled start time. Such transactions are not sent to
300 the server, but are skipped altogether and counted separately.
301
302 A high schedule lag time is an indication that the system cannot
303 process transactions at the specified rate, with the chosen number
304 of clients and threads. When the average transaction execution time
305 is longer than the scheduled interval between each transaction,
306 each successive transaction will fall further behind, and the
307 schedule lag time will keep increasing the longer the test run is.
308 When that happens, you will have to reduce the specified
309 transaction rate.
310
311 -s scale_factor
312 --scale=scale_factor
313 Report the specified scale factor in pgbench's output. With the
314 built-in tests, this is not necessary; the correct scale factor
315 will be detected by counting the number of rows in the
316 pgbench_branches table. However, when testing only custom
317 benchmarks (-f option), the scale factor will be reported as 1
318 unless this option is used.
319
320 -S
321 --select-only
322 Run built-in select-only script. Shorthand for -b select-only.
323
324 -t transactions
325 --transactions=transactions
326 Number of transactions each client runs. Default is 10.
327
328 -T seconds
329 --time=seconds
330 Run the test for this many seconds, rather than a fixed number of
331 transactions per client. -t and -T are mutually exclusive.
332
333 -v
334 --vacuum-all
335 Vacuum all four standard tables before running the test. With
336 neither -n nor -v, pgbench will vacuum the pgbench_tellers and
337 pgbench_branches tables, and will truncate pgbench_history.
338
339 --aggregate-interval=seconds
340 Length of aggregation interval (in seconds). May be used only with
341 -l option. With this option, the log contains per-interval summary
342 data, as described below.
343
344 --log-prefix=prefix
345 Set the filename prefix for the log files created by --log. The
346 default is pgbench_log.
347
348 --progress-timestamp
349 When showing progress (option -P), use a timestamp (Unix epoch)
350 instead of the number of seconds since the beginning of the run.
351 The unit is in seconds, with millisecond precision after the dot.
352 This helps compare logs generated by various tools.
353
354 --random-seed=SEED
355 Set random generator seed. Seeds the system random number
356 generator, which then produces a sequence of initial generator
357 states, one for each thread. Values for SEED may be: time (the
358 default, the seed is based on the current time), rand (use a strong
359 random source, failing if none is available), or an unsigned
360 decimal integer value. The random generator is invoked explicitly
361 from a pgbench script (random... functions) or implicitly (for
362 instance option --rate uses it to schedule transactions). When
363 explicitly set, the value used for seeding is shown on the
364 terminal. Any value allowed for SEED may also be provided through
365 the environment variable PGBENCH_RANDOM_SEED. To ensure that the
366 provided seed impacts all possible uses, put this option first or
367 use the environment variable.
368
369 Setting the seed explicitly allows to reproduce a pgbench run
370 exactly, as far as random numbers are concerned. As the random
371 state is managed per thread, this means the exact same pgbench run
372 for an identical invocation if there is one client per thread and
373 there are no external or data dependencies. From a statistical
374 viewpoint reproducing runs exactly is a bad idea because it can
375 hide the performance variability or improve performance unduly,
376 e.g. by hitting the same pages as a previous run. However, it may
377 also be of great help for debugging, for instance re-running a
378 tricky case which leads to an error. Use wisely.
379
380 --sampling-rate=rate
381 Sampling rate, used when writing data into the log, to reduce the
382 amount of log generated. If this option is given, only the
383 specified fraction of transactions are logged. 1.0 means all
384 transactions will be logged, 0.05 means only 5% of the transactions
385 will be logged.
386
387 Remember to take the sampling rate into account when processing the
388 log file. For example, when computing TPS values, you need to
389 multiply the numbers accordingly (e.g. with 0.01 sample rate,
390 you'll only get 1/100 of the actual TPS).
391
392 Common Options
393 pgbench accepts the following command-line common arguments:
394
395 -h hostname
396 --host=hostname
397 The database server's host name
398
399 -p port
400 --port=port
401 The database server's port number
402
403 -U login
404 --username=login
405 The user name to connect as
406
407 -V
408 --version
409 Print the pgbench version and exit.
410
411 -?
412 --help
413 Show help about pgbench command line arguments, and exit.
414
416 A successful run will exit with status 0. Exit status 1 indicates
417 static problems such as invalid command-line options. Errors during the
418 run such as database errors or problems in the script will result in
419 exit status 2. In the latter case, pgbench will print partial results.
420
422 What Is the “Transaction” Actually Performed in pgbench?
423 pgbench executes test scripts chosen randomly from a specified list.
424 They include built-in scripts with -b and user-provided custom scripts
425 with -f. Each script may be given a relative weight specified after a @
426 so as to change its drawing probability. The default weight is 1.
427 Scripts with a weight of 0 are ignored.
428
429 The default built-in transaction script (also invoked with -b
430 tpcb-like) issues seven commands per transaction over randomly chosen
431 aid, tid, bid and delta. The scenario is inspired by the TPC-B
432 benchmark, but is not actually TPC-B, hence the name.
433
434 1. BEGIN;
435
436 2. UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid
437 = :aid;
438
439 3. SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
440
441 4. UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid =
442 :tid;
443
444 5. UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid
445 = :bid;
446
447 6. INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES
448 (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
449
450 7. END;
451
452 If you select the simple-update built-in (also -N), steps 4 and 5
453 aren't included in the transaction. This will avoid update contention
454 on these tables, but it makes the test case even less like TPC-B.
455
456 If you select the select-only built-in (also -S), only the SELECT is
457 issued.
458
459 Custom Scripts
460 pgbench has support for running custom benchmark scenarios by replacing
461 the default transaction script (described above) with a transaction
462 script read from a file (-f option). In this case a “transaction”
463 counts as one execution of a script file.
464
465 A script file contains one or more SQL commands terminated by
466 semicolons. Empty lines and lines beginning with -- are ignored. Script
467 files can also contain “meta commands”, which are interpreted by
468 pgbench itself, as described below.
469
470 Note
471 Before PostgreSQL 9.6, SQL commands in script files were terminated
472 by newlines, and so they could not be continued across lines. Now a
473 semicolon is required to separate consecutive SQL commands (though
474 a SQL command does not need one if it is followed by a meta
475 command). If you need to create a script file that works with both
476 old and new versions of pgbench, be sure to write each SQL command
477 on a single line ending with a semicolon.
478
479 There is a simple variable-substitution facility for script files.
480 Variable names must consist of letters (including non-Latin letters),
481 digits, and underscores. Variables can be set by the command-line -D
482 option, explained above, or by the meta commands explained below. In
483 addition to any variables preset by -D command-line options, there are
484 a few variables that are preset automatically, listed in Table 257. A
485 value specified for these variables using -D takes precedence over the
486 automatic presets. Once set, a variable's value can be inserted into a
487 SQL command by writing :variablename. When running more than one client
488 session, each session has its own set of variables. pgbench supports
489 up to 255 variable uses in one statement.
490
491 Table 257. Automatic Variables
492 ┌─────────────┬────────────────────────────┐
493 │Variable │ Description │
494 ├─────────────┼────────────────────────────┤
495 │client_id │ unique number identifying │
496 │ │ the client session (starts │
497 │ │ from zero) │
498 ├─────────────┼────────────────────────────┤
499 │default_seed │ seed used in hash │
500 │ │ functions by default │
501 ├─────────────┼────────────────────────────┤
502 │random_seed │ random generator seed │
503 │ │ (unless overwritten with │
504 │ │ -D) │
505 ├─────────────┼────────────────────────────┤
506 │scale │ current scale factor │
507 └─────────────┴────────────────────────────┘
508
509 Script file meta commands begin with a backslash (\) and normally
510 extend to the end of the line, although they can be continued to
511 additional lines by writing backslash-return. Arguments to a meta
512 command are separated by white space. These meta commands are
513 supported:
514
515 \gset [prefix]
516 This command may be used to end SQL queries, taking the place of
517 the terminating semicolon (;).
518
519 When this command is used, the preceding SQL query is expected to
520 return one row, the columns of which are stored into variables
521 named after column names, and prefixed with prefix if provided.
522
523 The following example puts the final account balance from the first
524 query into variable abalance, and fills variables p_two and p_three
525 with integers from the third query. The result of the second query
526 is discarded.
527
528 UPDATE pgbench_accounts
529 SET abalance = abalance + :delta
530 WHERE aid = :aid
531 RETURNING abalance \gset
532 -- compound of two queries
533 SELECT 1 \;
534 SELECT 2 AS two, 3 AS three \gset p_
535
536
537 \if expression
538 \elif expression
539 \else
540 \endif
541 This group of commands implements nestable conditional blocks,
542 similarly to psql's \if expression. Conditional expressions are
543 identical to those with \set, with non-zero values interpreted as
544 true.
545
546 \set varname expression
547 Sets variable varname to a value calculated from expression. The
548 expression may contain the NULL constant, Boolean constants TRUE
549 and FALSE, integer constants such as 5432, double constants such as
550 3.14159, references to variables :variablename, operators with
551 their usual SQL precedence and associativity, function calls, SQL
552 CASE generic conditional expressions and parentheses.
553
554 Functions and most operators return NULL on NULL input.
555
556 For conditional purposes, non zero numerical values are TRUE, zero
557 numerical values and NULL are FALSE.
558
559 Too large or small integer and double constants, as well as integer
560 arithmetic operators (+, -, * and /) raise errors on overflows.
561
562 When no final ELSE clause is provided to a CASE, the default value
563 is NULL.
564
565 Examples:
566
567 \set ntellers 10 * :scale
568 \set aid (1021 * random(1, 100000 * :scale)) % \
569 (100000 * :scale) + 1
570 \set divx CASE WHEN :x <> 0 THEN :y/:x ELSE NULL END
571
572 \sleep number [ us | ms | s ]
573 Causes script execution to sleep for the specified duration in
574 microseconds (us), milliseconds (ms) or seconds (s). If the unit is
575 omitted then seconds are the default. number can be either an
576 integer constant or a :variablename reference to a variable having
577 an integer value.
578
579 Example:
580
581 \sleep 10 ms
582
583 \setshell varname command [ argument ... ]
584 Sets variable varname to the result of the shell command command
585 with the given argument(s). The command must return an integer
586 value through its standard output.
587
588 command and each argument can be either a text constant or a
589 :variablename reference to a variable. If you want to use an
590 argument starting with a colon, write an additional colon at the
591 beginning of argument.
592
593 Example:
594
595 \setshell variable_to_be_assigned command literal_argument :variable ::literal_starting_with_colon
596
597 \shell command [ argument ... ]
598 Same as \setshell, but the result of the command is discarded.
599
600 Example:
601
602 \shell command literal_argument :variable ::literal_starting_with_colon
603
604 Built-in Operators
605 The arithmetic, bitwise, comparison and logical operators listed in
606 Table 258 are built into pgbench and may be used in expressions
607 appearing in \set.
608
609 Table 258. pgbench Operators by Increasing Precedence
610 ┌──────────────────┬──────────────────┬───────────┬────────┐
611 │Operator │ Description │ Example │ Result │
612 ├──────────────────┼──────────────────┼───────────┼────────┤
613 │OR │ logical or │ 5 or 0 │ TRUE │
614 ├──────────────────┼──────────────────┼───────────┼────────┤
615 │AND │ logical and │ 3 and 0 │ FALSE │
616 ├──────────────────┼──────────────────┼───────────┼────────┤
617 │NOT │ logical not │ not false │ TRUE │
618 ├──────────────────┼──────────────────┼───────────┼────────┤
619 │IS [NOT] │ value tests │ 1 is null │ FALSE │
620 │(NULL|TRUE|FALSE) │ │ │ │
621 ├──────────────────┼──────────────────┼───────────┼────────┤
622 │ISNULL|NOTNULL │ null tests │ 1 notnull │ TRUE │
623 ├──────────────────┼──────────────────┼───────────┼────────┤
624 │= │ is equal │ 5 = 4 │ FALSE │
625 ├──────────────────┼──────────────────┼───────────┼────────┤
626 │<> │ is not equal │ 5 <> 4 │ TRUE │
627 ├──────────────────┼──────────────────┼───────────┼────────┤
628 │!= │ is not equal │ 5 != 5 │ FALSE │
629 ├──────────────────┼──────────────────┼───────────┼────────┤
630 │< │ lower than │ 5 < 4 │ FALSE │
631 ├──────────────────┼──────────────────┼───────────┼────────┤
632 │<= │ lower or equal │ 5 <= 4 │ FALSE │
633 ├──────────────────┼──────────────────┼───────────┼────────┤
634 │> │ greater than │ 5 > 4 │ TRUE │
635 ├──────────────────┼──────────────────┼───────────┼────────┤
636 │>= │ greater or equal │ 5 >= 4 │ TRUE │
637 ├──────────────────┼──────────────────┼───────────┼────────┤
638 │| │ integer bitwise │ 1 | 2 │ 3 │
639 │ │ OR │ │ │
640 ├──────────────────┼──────────────────┼───────────┼────────┤
641 │# │ integer bitwise │ 1 # 3 │ 2 │
642 │ │ XOR │ │ │
643 ├──────────────────┼──────────────────┼───────────┼────────┤
644 │& │ integer bitwise │ 1 & 3 │ 1 │
645 │ │ AND │ │ │
646 ├──────────────────┼──────────────────┼───────────┼────────┤
647 │~ │ integer bitwise │ ~ 1 │ -2 │
648 │ │ NOT │ │ │
649 ├──────────────────┼──────────────────┼───────────┼────────┤
650 │<< │ integer bitwise │ 1 << 2 │ 4 │
651 │ │ shift left │ │ │
652 ├──────────────────┼──────────────────┼───────────┼────────┤
653 │>> │ integer bitwise │ 8 >> 2 │ 2 │
654 │ │ shift right │ │ │
655 ├──────────────────┼──────────────────┼───────────┼────────┤
656 │+ │ addition │ 5 + 4 │ 9 │
657 ├──────────────────┼──────────────────┼───────────┼────────┤
658 │- │ subtraction │ 3 - 2.0 │ 1.0 │
659 ├──────────────────┼──────────────────┼───────────┼────────┤
660 │* │ multiplication │ 5 * 4 │ 20 │
661 ├──────────────────┼──────────────────┼───────────┼────────┤
662 │/ │ division │ 5 / 3 │ 1 │
663 │ │ (integer │ │ │
664 │ │ truncates the │ │ │
665 │ │ results) │ │ │
666 ├──────────────────┼──────────────────┼───────────┼────────┤
667 │% │ modulo │ 3 % 2 │ 1 │
668 ├──────────────────┼──────────────────┼───────────┼────────┤
669 │- │ opposite │ - 2.0 │ -2.0 │
670 └──────────────────┴──────────────────┴───────────┴────────┘
671
672 Built-In Functions
673 The functions listed in Table 259 are built into pgbench and may be
674 used in expressions appearing in \set.
675
676 Table 259. pgbench Functions
677 ┌───────────────────────┬───────────────┬───────────────────────────┬───────────────────────┬────────────────────────┐
678 │Function │ Return Type │ Description │ Example │ Result │
679 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
680 │abs(a) │ same as a │ absolute │ abs(-17) │ 17 │
681 │ │ │ value │ │ │
682 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
683 │debug(a) │ same as a │ print a to │ debug(5432.1) │ 5432.1 │
684 │ │ │ stderr, │ │ │
685 │ │ │ and │ │ │
686 │ │ │ return a │ │ │
687 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
688 │double(i) │ double │ cast to │ double(5432) │ 5432.0 │
689 │ │ │ double │ │ │
690 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
691 │exp(x) │ double │ exponential │ exp(1.0) │ 2.718281828459045 │
692 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
693 │greatest(a [, │ double if any │ largest value │ greatest(5, │ 5 │
694 │... ] ) │ a is double, │ among │ 4, 3, 2) │ │
695 │ │ else integer │ arguments │ │ │
696 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
697 │hash(a [, │ integer │ alias for │ hash(10, │ -5817877081768721676 │
698 │seed ] ) │ │ hash_murmur2() │ 5432) │ │
699 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
700 │hash_fnv1a(a │ integer │ FNV-1a hash │ hash_fnv1a(10, │ -7793829335365542153 │
701 │[, seed ] ) │ │ │ 5432) │ │
702 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
703 │hash_murmur2(a │ integer │ MurmurHash2 │ hash_murmur2(10, │ -5817877081768721676 │
704 │[, seed ] ) │ │ hash │ 5432) │ │
705 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
706 │int(x) │ integer │ cast to int │ int(5.4 + 3.8) │ 9 │
707 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
708 │least(a [, ... │ double if any │ smallest value │ least(5, 4, 3, │ 2.1 │
709 │] ) │ a is double, │ among │ 2.1) │ │
710 │ │ else integer │ arguments │ │ │
711 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
712 │ln(x) │ double │ natural │ ln(2.718281828459045) │ 1.0 │
713 │ │ │ logarithm │ │ │
714 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
715 │mod(i, j) │ integer │ modulo │ mod(54, 32) │ 22 │
716 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
717 │pi() │ double │ value of the │ pi() │ 3.14159265358979323846 │
718 │ │ │ constant PI │ │ │
719 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
720 │pow(x, y), │ double │ exponentiation │ pow(2.0, 10), │ 1024.0 │
721 │power(x, y) │ │ │ power(2.0, 10) │ │
722 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
723 │random(lb, ub) │ integer │ uniformly-distributed │ random(1, 10) │ an integer between 1 │
724 │ │ │ random integer │ │ and 10 │
725 │ │ │ in [lb, ub] │ │ │
726 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
727 │random_exponential(lb, │ integer │ exponentially-distributed │ random_exponential(1, │ an integer between 1 │
728 │ub, parameter) │ │ random integer in │ 10, 3.0) │ and 10 │
729 │ │ │ [lb, ub], │ │ │
730 │ │ │ see │ │ │
731 │ │ │ below │ │ │
732 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
733 │random_gaussian(lb, │ integer │ Gaussian-distributed │ random_gaussian(1, │ an integer between 1 │
734 │ub, parameter) │ │ random integer in [lb, │ 10, 2.5) │ and 10 │
735 │ │ │ ub], │ │ │
736 │ │ │ see below │ │ │
737 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
738 │random_zipfian(lb, ub, │ integer │ Zipfian-distributed │ random_zipfian(1, 10, │ an integer between 1 │
739 │parameter) │ │ random integer in [lb, │ 1.5) │ and 10 │
740 │ │ │ ub], │ │ │
741 │ │ │ see below │ │ │
742 ├───────────────────────┼───────────────┼───────────────────────────┼───────────────────────┼────────────────────────┤
743 │sqrt(x) │ double │ square root │ sqrt(2.0) │ 1.414213562 │
744 └───────────────────────┴───────────────┴───────────────────────────┴───────────────────────┴────────────────────────┘
745
746 The random function generates values using a uniform distribution, that
747 is all the values are drawn within the specified range with equal
748 probability. The random_exponential, random_gaussian and random_zipfian
749 functions require an additional double parameter which determines the
750 precise shape of the distribution.
751
752 · For an exponential distribution, parameter controls the
753 distribution by truncating a quickly-decreasing exponential
754 distribution at parameter, and then projecting onto integers
755 between the bounds. To be precise, with
756
757 f(x) = exp(-parameter * (x - min) / (max - min + 1)) / (1 - exp(-parameter))
758
759 Then value i between min and max inclusive is drawn with
760 probability: f(i) - f(i + 1).
761
762 Intuitively, the larger the parameter, the more frequently values
763 close to min are accessed, and the less frequently values close to
764 max are accessed. The closer to 0 parameter is, the flatter (more
765 uniform) the access distribution. A crude approximation of the
766 distribution is that the most frequent 1% values in the range,
767 close to min, are drawn parameter% of the time. The parameter value
768 must be strictly positive.
769
770 · For a Gaussian distribution, the interval is mapped onto a standard
771 normal distribution (the classical bell-shaped Gaussian curve)
772 truncated at -parameter on the left and +parameter on the right.
773 Values in the middle of the interval are more likely to be drawn.
774 To be precise, if PHI(x) is the cumulative distribution function of
775 the standard normal distribution, with mean mu defined as (max +
776 min) / 2.0, with
777
778 f(x) = PHI(2.0 * parameter * (x - mu) / (max - min + 1)) /
779 (2.0 * PHI(parameter) - 1)
780
781 then value i between min and max inclusive is drawn with
782 probability: f(i + 0.5) - f(i - 0.5). Intuitively, the larger the
783 parameter, the more frequently values close to the middle of the
784 interval are drawn, and the less frequently values close to the min
785 and max bounds. About 67% of values are drawn from the middle 1.0 /
786 parameter, that is a relative 0.5 / parameter around the mean, and
787 95% in the middle 2.0 / parameter, that is a relative 1.0 /
788 parameter around the mean; for instance, if parameter is 4.0, 67%
789 of values are drawn from the middle quarter (1.0 / 4.0) of the
790 interval (i.e. from 3.0 / 8.0 to 5.0 / 8.0) and 95% from the middle
791 half (2.0 / 4.0) of the interval (second and third quartiles). The
792 minimum allowed parameter value is 2.0.
793
794 · random_zipfian generates a bounded Zipfian distribution. parameter
795 defines how skewed the distribution is. The larger the parameter,
796 the more frequently values closer to the beginning of the interval
797 are drawn. The distribution is such that, assuming the range starts
798 from 1, the ratio of the probability of drawing k versus drawing
799 k+1 is ((k+1)/k)**parameter. For example, random_zipfian(1, ...,
800 2.5) produces the value 1 about (2/1)**2.5 = 5.66 times more
801 frequently than 2, which itself is produced (3/2)**2.5 = 2.76 times
802 more frequently than 3, and so on.
803
804 pgbench's implementation is based on "Non-Uniform Random Variate
805 Generation", Luc Devroye, p. 550-551, Springer 1986. Due to
806 limitations of that algorithm, the parameter value is restricted to
807 the range [1.001, 1000].
808
809 Hash functions hash, hash_murmur2 and hash_fnv1a accept an input value
810 and an optional seed parameter. In case the seed isn't provided the
811 value of :default_seed is used, which is initialized randomly unless
812 set by the command-line -D option. Hash functions can be used to
813 scatter the distribution of random functions such as random_zipfian or
814 random_exponential. For instance, the following pgbench script
815 simulates possible real world workload typical for social media and
816 blogging platforms where few accounts generate excessive load:
817
818 \set r random_zipfian(0, 100000000, 1.07)
819 \set k abs(hash(:r)) % 1000000
820
821 In some cases several distinct distributions are needed which don't
822 correlate with each other and this is when implicit seed parameter
823 comes in handy:
824
825 \set k1 abs(hash(:r, :default_seed + 123)) % 1000000
826 \set k2 abs(hash(:r, :default_seed + 321)) % 1000000
827
828 As an example, the full definition of the built-in TPC-B-like
829 transaction is:
830
831 \set aid random(1, 100000 * :scale)
832 \set bid random(1, 1 * :scale)
833 \set tid random(1, 10 * :scale)
834 \set delta random(-5000, 5000)
835 BEGIN;
836 UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
837 SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
838 UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
839 UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
840 INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
841 END;
842
843 This script allows each iteration of the transaction to reference
844 different, randomly-chosen rows. (This example also shows why it's
845 important for each client session to have its own variables — otherwise
846 they'd not be independently touching different rows.)
847
848 Per-Transaction Logging
849 With the -l option (but without the --aggregate-interval option),
850 pgbench writes information about each transaction to a log file. The
851 log file will be named prefix.nnn, where prefix defaults to
852 pgbench_log, and nnn is the PID of the pgbench process. The prefix can
853 be changed by using the --log-prefix option. If the -j option is 2 or
854 higher, so that there are multiple worker threads, each will have its
855 own log file. The first worker will use the same name for its log file
856 as in the standard single worker case. The additional log files for the
857 other workers will be named prefix.nnn.mmm, where mmm is a sequential
858 number for each worker starting with 1.
859
860 The format of the log is:
861
862 client_id transaction_no time script_no time_epoch time_us [ schedule_lag ]
863
864 where client_id indicates which client session ran the transaction,
865 transaction_no counts how many transactions have been run by that
866 session, time is the total elapsed transaction time in microseconds,
867 script_no identifies which script file was used (useful when multiple
868 scripts were specified with -f or -b), and time_epoch/time_us are a
869 Unix-epoch time stamp and an offset in microseconds (suitable for
870 creating an ISO 8601 time stamp with fractional seconds) showing when
871 the transaction completed. The schedule_lag field is the difference
872 between the transaction's scheduled start time, and the time it
873 actually started, in microseconds. It is only present when the --rate
874 option is used. When both --rate and --latency-limit are used, the time
875 for a skipped transaction will be reported as skipped.
876
877 Here is a snippet of a log file generated in a single-client run:
878
879 0 199 2241 0 1175850568 995598
880 0 200 2465 0 1175850568 998079
881 0 201 2513 0 1175850569 608
882 0 202 2038 0 1175850569 2663
883
884 Another example with --rate=100 and --latency-limit=5 (note the
885 additional schedule_lag column):
886
887 0 81 4621 0 1412881037 912698 3005
888 0 82 6173 0 1412881037 914578 4304
889 0 83 skipped 0 1412881037 914578 5217
890 0 83 skipped 0 1412881037 914578 5099
891 0 83 4722 0 1412881037 916203 3108
892 0 84 4142 0 1412881037 918023 2333
893 0 85 2465 0 1412881037 919759 740
894
895 In this example, transaction 82 was late, because its latency (6.173
896 ms) was over the 5 ms limit. The next two transactions were skipped,
897 because they were already late before they were even started.
898
899 When running a long test on hardware that can handle a lot of
900 transactions, the log files can become very large. The --sampling-rate
901 option can be used to log only a random sample of transactions.
902
903 Aggregated Logging
904 With the --aggregate-interval option, a different format is used for
905 the log files:
906
907 interval_start num_transactions sum_latency sum_latency_2 min_latency max_latency [ sum_lag sum_lag_2 min_lag max_lag [ skipped ] ]
908
909 where interval_start is the start of the interval (as a Unix epoch time
910 stamp), num_transactions is the number of transactions within the
911 interval, sum_latency is the sum of the transaction latencies within
912 the interval, sum_latency_2 is the sum of squares of the transaction
913 latencies within the interval, min_latency is the minimum latency
914 within the interval, and max_latency is the maximum latency within the
915 interval. The next fields, sum_lag, sum_lag_2, min_lag, and max_lag,
916 are only present if the --rate option is used. They provide statistics
917 about the time each transaction had to wait for the previous one to
918 finish, i.e. the difference between each transaction's scheduled start
919 time and the time it actually started. The very last field, skipped, is
920 only present if the --latency-limit option is used, too. It counts the
921 number of transactions skipped because they would have started too
922 late. Each transaction is counted in the interval when it was
923 committed.
924
925 Here is some example output:
926
927 1345828501 5601 1542744 483552416 61 2573
928 1345828503 7884 1979812 565806736 60 1479
929 1345828505 7208 1979422 567277552 59 1391
930 1345828507 7685 1980268 569784714 60 1398
931 1345828509 7073 1979779 573489941 236 1411
932
933 Notice that while the plain (unaggregated) log file shows which script
934 was used for each transaction, the aggregated log does not. Therefore
935 if you need per-script data, you need to aggregate the data on your
936 own.
937
938 Per-Statement Latencies
939 With the -r option, pgbench collects the elapsed transaction time of
940 each statement executed by every client. It then reports an average of
941 those values, referred to as the latency for each statement, after the
942 benchmark has finished.
943
944 For the default script, the output will look similar to this:
945
946 starting vacuum...end.
947 transaction type: <builtin: TPC-B (sort of)>
948 scaling factor: 1
949 query mode: simple
950 number of clients: 10
951 number of threads: 1
952 number of transactions per client: 1000
953 number of transactions actually processed: 10000/10000
954 latency average = 15.844 ms
955 latency stddev = 2.715 ms
956 tps = 618.764555 (including connections establishing)
957 tps = 622.977698 (excluding connections establishing)
958 statement latencies in milliseconds:
959 0.002 \set aid random(1, 100000 * :scale)
960 0.005 \set bid random(1, 1 * :scale)
961 0.002 \set tid random(1, 10 * :scale)
962 0.001 \set delta random(-5000, 5000)
963 0.326 BEGIN;
964 0.603 UPDATE pgbench_accounts SET abalance = abalance + :delta WHERE aid = :aid;
965 0.454 SELECT abalance FROM pgbench_accounts WHERE aid = :aid;
966 5.528 UPDATE pgbench_tellers SET tbalance = tbalance + :delta WHERE tid = :tid;
967 7.335 UPDATE pgbench_branches SET bbalance = bbalance + :delta WHERE bid = :bid;
968 0.371 INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (:tid, :bid, :aid, :delta, CURRENT_TIMESTAMP);
969 1.212 END;
970
971 If multiple script files are specified, the averages are reported
972 separately for each script file.
973
974 Note that collecting the additional timing information needed for
975 per-statement latency computation adds some overhead. This will slow
976 average execution speed and lower the computed TPS. The amount of
977 slowdown varies significantly depending on platform and hardware.
978 Comparing average TPS values with and without latency reporting enabled
979 is a good way to measure if the timing overhead is significant.
980
981 Good Practices
982 It is very easy to use pgbench to produce completely meaningless
983 numbers. Here are some guidelines to help you get useful results.
984
985 In the first place, never believe any test that runs for only a few
986 seconds. Use the -t or -T option to make the run last at least a few
987 minutes, so as to average out noise. In some cases you could need hours
988 to get numbers that are reproducible. It's a good idea to try the test
989 run a few times, to find out if your numbers are reproducible or not.
990
991 For the default TPC-B-like test scenario, the initialization scale
992 factor (-s) should be at least as large as the largest number of
993 clients you intend to test (-c); else you'll mostly be measuring update
994 contention. There are only -s rows in the pgbench_branches table, and
995 every transaction wants to update one of them, so -c values in excess
996 of -s will undoubtedly result in lots of transactions blocked waiting
997 for other transactions.
998
999 The default test scenario is also quite sensitive to how long it's been
1000 since the tables were initialized: accumulation of dead rows and dead
1001 space in the tables changes the results. To understand the results you
1002 must keep track of the total number of updates and when vacuuming
1003 happens. If autovacuum is enabled it can result in unpredictable
1004 changes in measured performance.
1005
1006 A limitation of pgbench is that it can itself become the bottleneck
1007 when trying to test a large number of client sessions. This can be
1008 alleviated by running pgbench on a different machine from the database
1009 server, although low network latency will be essential. It might even
1010 be useful to run several pgbench instances concurrently, on several
1011 client machines, against the same database server.
1012
1013 Security
1014 If untrusted users have access to a database that has not adopted a
1015 secure schema usage pattern, do not run pgbench in that database.
1016 pgbench uses unqualified names and does not manipulate the search path.
1017
1018
1019
1020PostgreSQL 12.2 2020 PGBENCH(1)