Merge build trunk.
[awesomized/libmemcached] / docs / man / memaslap.1
1 .TH "MEMASLAP" "1" "October 18, 2011" "1.01" "libmemcached"
2 .SH NAME
3 memaslap \- libmemcached Documentation
4 .
5 .nr rst2man-indent-level 0
6 .
7 .de1 rstReportMargin
8 \\$1 \\n[an-margin]
9 level \\n[rst2man-indent-level]
10 level margin: \\n[rst2man-indent\\n[rst2man-indent-level]]
11 -
12 \\n[rst2man-indent0]
13 \\n[rst2man-indent1]
14 \\n[rst2man-indent2]
15 ..
16 .de1 INDENT
17 .\" .rstReportMargin pre:
18 . RS \\$1
19 . nr rst2man-indent\\n[rst2man-indent-level] \\n[an-margin]
20 . nr rst2man-indent-level +1
21 .\" .rstReportMargin post:
22 ..
23 .de UNINDENT
24 . RE
25 .\" indent \\n[an-margin]
26 .\" old: \\n[rst2man-indent\\n[rst2man-indent-level]]
27 .nr rst2man-indent-level -1
28 .\" new: \\n[rst2man-indent\\n[rst2man-indent-level]]
29 .in \\n[rst2man-indent\\n[rst2man-indent-level]]u
30 ..
31 .\" Man page generated from reStructeredText.
32 .
33 .SH SYNOPSIS
34 .sp
35 memaslap [options]
36 .INDENT 0.0
37 .TP
38 .B \-\-help
39 .UNINDENT
40 .SH DESCRIPTION
41 .sp
42 \fBmemaslap\fP is a load generation and benchmark tool for memcached
43 servers. It generates configurable workload such as threads, concurrencies,
44 connections, run time, overwrite, miss rate, key size, value size, get/set
45 proportion, expected throughput, and so on. Furthermore, it also testss data
46 verification, expire\-time verification, UDP, binary protocol, facebook test,
47 replication test, multi\-get and reconnection, etc.
48 .sp
49 Memaslap manages network connections like memcached with
50 libevent. Each thread of memaslap is bound with a CPU core, all
51 the threads don\(aqt communicate with each other, and there are several socket
52 connections in each thread. Each connection keeps key size distribution,
53 value size distribution, and command distribution by itself.
54 .sp
55 You can specify servers via the \fI\-\-servers\fP option or via the
56 environment variable \fBMEMCACHED_SERVERS\fP.
57 .SH FEATURES
58 .sp
59 Memslap is developed to for the following purposes:
60 .sp
61 Manages network connections with libevent asynchronously.
62 .sp
63 Set both TCP and UDP up to use non\-blocking IO.
64 .sp
65 Improves parallelism: higher performance in multi\-threads environments.
66 .sp
67 Improves time efficiency: faster processing speed.
68 .sp
69 Generates key and value more efficiently; key size distribution and value size distribution are configurable.
70 .sp
71 Supports get, multi\-get, and set commands; command distribution is configurable.
72 .sp
73 Supports controllable miss rate and overwrite rate.
74 .sp
75 Supports data and expire\-time verification.
76 .sp
77 Supports dumping statistic information periodically.
78 .sp
79 Supports thousands of TCP connections.
80 .sp
81 Supports binary protocol.
82 .sp
83 Supports facebook test (set with TCP and multi\-get with UDP) and replication test.
84 .SH DETAILS
85 .SS Effective implementation of network.
86 .sp
87 For memaslap, both TCP and UDP use non\-blocking network IO. All
88 the network events are managed by libevent as memcached. The network module
89 of memaslap is similar to memcached. Libevent can ensure
90 memaslap can handle network very efficiently.
91 .SS Effective implementation of multi\-threads and concurrency
92 .sp
93 Memslap has the similar implementation of multi\-threads to
94 memcached. Memslap creates one or more self\-governed threads;
95 each thread is bound with one CPU core if the system testss setting CPU
96 core affinity.
97 .sp
98 In addition, each thread has a libevent to manage the events of the network;
99 each thread has one or more self\-governed concurrencies; and each
100 concurrency has one or more socket connections. All the concurrencies don’t
101 communicate with each other even though they are in the same thread.
102 .sp
103 Memslap can create thousands of socket connections, and each
104 concurrency has tens of socket connections. Each concurrency randomly or
105 sequentially selects one socket connection from its socket connection pool
106 to run, so memaslap can ensure each concurrency handles one
107 socket connection at any given time. Users can specify the number of
108 concurrency and socket connections of each concurrency according to their
109 expected workload.
110 .SS Effective implementation of generating key and value
111 .sp
112 In order to improve time efficiency and space efficiency,
113 memaslap creates a random characters table with 10M characters. All the
114 suffixes of keys and values are generated from this random characters table.
115 .sp
116 Memslap uses the offset in the character table and the length
117 of the string to identify a string. It can save much memory.
118 Each key contains two parts, a prefix and a suffix. The prefix is an
119 uint64_t, 8 bytes. In order to verify the data set before,
120 memaslap need to ensure each key is unique, so it uses the prefix to identify
121 a key. The prefix cannot include illegal characters, such as ‘r’, ‘n’,
122 ‘0’ and ‘ ‘. And memaslap has an algorithm to ensure that.
123 .sp
124 Memslap doesn’t generate all the objects (key\-value pairs) at
125 the beginning. It only generates enough objects to fill the task window
126 (default 10K objects) of each concurrency. Each object has the following
127 basic information, key prefix, key suffix offset in the character table, key
128 length, value offset in the character table, and value length.
129 .sp
130 In the work process, each concurrency sequentially or randomly selects an
131 object from the window to do set operation or get operation. At the same
132 time, each concurrency kicks objects out of its window and adds new object
133 into it.
134 .SS Simple but useful task scheduling
135 .sp
136 Memslap uses libevent to schedule all the concurrencies of
137 threads, and each concurrency schedules tasks based on the local task
138 window. Memslap assumes that if each concurrency keeps the same
139 key distribution, value distribution and commands distribution, from
140 outside, memaslap keeps all the distribution as a whole.
141 Each task window includes a lot of objects, each object stores its basic
142 information, such as key, value, expire time, and so on. At any time, all
143 the objects in the window keep the same and fixed key and value
144 distribution. If an object is overwritten, the value of the object will be
145 updated. Memslap verifies the data or expire\-time according to
146 the object information stored in the task window.
147 .sp
148 Libevent selects which concurrency to handle based on a specific network
149 event. Then the concurrency selects which command (get or set) to operate
150 based on the command distribution. If it needs to kick out an old object and
151 add a new object, in order to keep the same key and value distribution, the
152 new object must have the same key length and value length.
153 .sp
154 If memcached server has two cache layers (memory and SSD), running
155 memaslap with different window sizes can get different cache
156 miss rates. If memaslap adds enough objects into the windows at
157 the beginning, and the cache of memcached cannot store all the objects
158 initialized, then memaslap will get some objects from the second
159 cache layer. It causes the first cache layer to miss. So the user can
160 specify the window size to get the expected miss rate of the first cache
161 layer.
162 .SS Useful implementation of multi\-servers , UDP, TCP, multi\-get and binary protocol
163 .sp
164 Because each thread is self\-governed, memaslap can assign
165 different threads to handle different memcached servers. This is just one of
166 the ways in which memaslap tests multiple servers. The only
167 limitation is that the number of servers cannot be greater than the number
168 of threads. The other way to test multiple servers is for replication
169 test. Each concurrency has one socket connection to each memcached server.
170 For the implementation, memaslap can set some objects to one
171 memcached server, and get these objects from the other servers.
172 .sp
173 By default, Memslap does single get. If the user specifies
174 multi\-get option, memaslap will collect enough get commands and
175 pack and send the commands together.
176 .sp
177 Memslap testss both the ASCII protocol and binary protocol,
178 but it runs on the ASCII protocol by default.
179 Memslap by default runs on the TCP protocol, but it also
180 tests UDP. Because UDP is unreliable, dropped packages and out\-of\-order
181 packages may occur. Memslap creates a memory buffer to handle
182 these problems. Memslap tries to read all the response data of
183 one command from the server and reorders the response data. If some packages
184 get lost, the waiting timeout mechanism can ensure half\-baked packages will
185 be discarded and the next command will be sent.
186 .SH USAGE
187 .sp
188 Below are some usage samples:
189 .sp
190 memaslap \-s 127.0.0.1:11211 \-S 5s
191 .sp
192 memaslap \-s 127.0.0.1:11211 \-t 2m \-v 0.2 \-e 0.05 \-b
193 .sp
194 memaslap \-s 127.0.0.1:11211 \-F config \-t 2m \-w 40k \-S 20s \-o 0.2
195 .sp
196 memaslap \-s 127.0.0.1:11211 \-F config \-t 2m \-T 4 \-c 128 \-d 20 \-P 40k
197 .sp
198 memaslap \-s 127.0.0.1:11211 \-F config \-t 2m \-d 50 \-a \-n 40
199 .sp
200 memaslap \-s 127.0.0.1:11211,127.0.0.1:11212 \-F config \-t 2m
201 .sp
202 memaslap \-s 127.0.0.1:11211,127.0.0.1:11212 \-F config \-t 2m \-p 2
203 .sp
204 The user must specify one server at least to run memaslap. The
205 rest of the parameters have default values, as shown below:
206 .sp
207 Thread number = 1 Concurrency = 16
208 .sp
209 Run time = 600 seconds Configuration file = NULL
210 .sp
211 Key size = 64 Value size = 1024
212 .sp
213 Get/set = 9:1 Window size = 10k
214 .sp
215 Execute number = 0 Single get = true
216 .sp
217 Multi\-get = false Number of sockets of each concurrency = 1
218 .sp
219 Reconnect = false Data verification = false
220 .sp
221 Expire\-time verification = false ASCII protocol = true
222 .sp
223 Binary protocol = false Dumping statistic information
224 .sp
225 periodically = false
226 .sp
227 Overwrite proportion = 0% UDP = false
228 .sp
229 TCP = true Limit throughput = false
230 .sp
231 Facebook test = false Replication test = false
232 .SS Key size, value size and command distribution.
233 .sp
234 All the distributions are read from the configuration file specified by user
235 with “—cfg_cmd” option. If the user does not specify a configuration file,
236 memaslap will run with the default distribution (key size = 64,
237 value size = 1024, get/set = 9:1). For information on how to edit the
238 configuration file, refer to the “Configuration File” section.
239 .sp
240 The minimum key size is 16 bytes; the maximum key size is 250 bytes. The
241 precision of proportion is 0.001. The proportion of distribution will be
242 rounded to 3 decimal places.
243 .sp
244 The minimum value size is 1 bytes; the maximum value size is 1M bytes. The
245 precision of proportion is 0.001. The proportion of distribution will be
246 rounded to 3 decimal places.
247 Currently, memaslap only testss set and get commands. And it
248 testss 100% set and 100% get. For 100% get, it will preset some objects to
249 the server.
250 .SS Multi\-thread and concurrency
251 .sp
252 The high performance of memaslap benefits from the special
253 schedule of thread and concurrency. It’s important to specify the proper
254 number of them. The default number of threads is 1; the default number of
255 concurrency is 16. The user can use “—threads” and “\-\-concurrency” to
256 specify these variables.
257 .sp
258 If the system tests setting CPU affinity and the number of threads
259 specified by the user is greater than 1, memaslap will try to
260 bind each thread to a different CPU core. So if you want to get the best
261 performance memaslap, it is better to specify the number of
262 thread equal to the number of CPU cores. The number of threads specified by
263 the user can also be less or greater than the number of CPU cores. Because
264 of the limitation of implementation, the number of concurrencies could be
265 the multiple of the number of threads.
266 .INDENT 0.0
267 .IP 1. 3
268 .
269 For 8 CPU cores system
270 .UNINDENT
271 .sp
272 For example:
273 .sp
274 \-\-threads=2 \-\-concurrency=128
275 .sp
276 \-\-threads=8 \-\-concurrency=128
277 .sp
278 \-\-threads=8 \-\-concurrency=256
279 .sp
280 \-\-threads=12 \-\-concurrency=144
281 .INDENT 0.0
282 .IP 2. 3
283 .
284 For 16 CPU cores system
285 .UNINDENT
286 .sp
287 For example:
288 .sp
289 \-\-threads=8 \-\-concurrency=128
290 .sp
291 \-\-threads=16 \-\-concurrency=256
292 .sp
293 \-\-threads=16 \-\-concurrency=512
294 .sp
295 \-\-threads=24 \-\-concurrency=288
296 .sp
297 The memaslap performs very well, when
298 used to test the performance of memcached servers.
299 Most of the time, the bottleneck is the network or
300 the server. If for some reason the user wants to
301 limit the performance of memaslap, there
302 are two ways to do this:
303 .sp
304 Decrease the number of threads and concurrencies.
305 Use the option “\-\-tps” that memaslap
306 provides to limit the throughput. This option allows
307 the user to get the expected throughput. For
308 example, assume that the maximum throughput is 50
309 kops/s for a specific configuration, you can specify
310 the throughput equal to or less than the maximum
311 throughput using “\-\-tps” option.
312 .SS Window size
313 .sp
314 Most of the time, the user does not need to specify the window size. The
315 default window size is 10k. For Schooner Memcached, the user can specify
316 different window sizes to get different cache miss rates based on the test
317 case. Memslap testss cache miss rate between 0% and 100%.
318 If you use this utility to test the performance of Schooner Memcached, you
319 can specify a proper window size to get the expected cache miss rate. The
320 formula for calculating window size is as follows:
321 .sp
322 Assume that the key size is 128 bytes, and the value size is 2048 bytes, and
323 concurrency=128.
324 .sp
325 1. Small cache cache_size=1M, 100% cache miss (all data get from SSD).
326 win_size=10k
327 .INDENT 0.0
328 .IP 2. 3
329 .
330 cache_size=4G
331 .UNINDENT
332 .sp
333 (1). cache miss rate 0%
334 .sp
335 win_size=8k
336 .sp
337 (2). cache miss rate 5%
338 .sp
339 win_size=11k
340 .INDENT 0.0
341 .IP 3. 3
342 .
343 cache_size=16G
344 .UNINDENT
345 .sp
346 (1). cache miss rate 0%
347 .sp
348 win_size=32k
349 .sp
350 (2). cache miss
351 .sp
352 rate 5%
353 .sp
354 win_size=46k
355 .sp
356 The formula for calculating window size for cache miss rate 0%:
357 .sp
358 cache_size / concurrency / (key_size + value_size) * 0.5
359 .sp
360 The formula for calculating window size for cache miss rate 5%:
361 .sp
362 cache_size / concurrency / (key_size + value_size) * 0.7
363 .SS Verification
364 .sp
365 Memslap testss both data verification and expire\-time
366 verification. The user can use "\-\-verify=" or "\-v" to specify the proportion
367 of data verification. In theory, it testss 100% data verification. The
368 user can use "\-\-exp_verify=" or "\-e" to specify the proportion of
369 expire\-time verification. In theory, it testss 100% expire\-time
370 verification. Specify the "\-\-verbose" options to get more detailed error
371 information.
372 .sp
373 For example: \-\-exp_verify=0.01 –verify=0.1 , it means that 1% of the objects
374 set with expire\-time, 10% of the objects gotten will be verified. If the
375 objects are gotten, memaslap will verify the expire\-time and
376 value.
377 .SS multi\-servers and multi\-config
378 .sp
379 Memslap testss multi\-servers based on self\-governed thread.
380 There is a limitation that the number of servers cannot be greater than the
381 number of threads. Memslap assigns one thread to handle one
382 server at least. The user can use the "\-\-servers=" or "\-s" option to specify
383 multi\-servers.
384 .sp
385 For example:
386 .sp
387 \-\-servers=10.1.1.1:11211,10.1.1.2:11212,10.1.1.3:11213 \-\-threads=6 \-\-concurrency=36
388 .sp
389 The above command means that there are 6 threads, with each thread having 6
390 concurrencies and that threads 0 and 3 handle server 0 (10.1.1.1); threads 1
391 and 4 handle server 1 (10.1.1.2); and thread 2 and 5 handle server 2
392 (10.1.1.3).
393 .sp
394 All the threads and concurrencies in memaslap are self\-governed.
395 .sp
396 So is memaslap. The user can start up several
397 memaslap instances. The user can run memaslap on different client
398 machines to communicate with the same memcached server at the same. It is
399 recommended that the user start different memaslap on different
400 machines using the same configuration.
401 .SS Run with execute number mode or time mode
402 .sp
403 The default memaslap runs with time mode. The default run time
404 is 10 minutes. If it times out, memaslap will exit. Do not
405 specify both execute number mode and time mode at the same time; just
406 specify one instead.
407 .sp
408 For example:
409 .sp
410 \-\-time=30s (It means the test will run 30 seconds.)
411 .sp
412 \-\-execute_number=100000 (It means that after running 100000 commands, the test will exit.)
413 .SS Dump statistic information periodically.
414 .sp
415 The user can use "\-\-stat_freq=" or "\-S" to specify the frequency.
416 .sp
417 For example:
418 .sp
419 \-\-stat_freq=20s
420 .sp
421 Memslap will dump the statistics of the commands (get and set) at the frequency of every 20
422 seconds.
423 .sp
424 For more information on the format of dumping statistic information, refer to “Format of Output” section.
425 .SS Multi\-get
426 .sp
427 The user can use "\-\-division=" or "\-d" to specify multi\-get keys count.
428 Memslap by default does single get with TCP. Memslap also testss data
429 verification and expire\-time verification for multi\-get.
430 .sp
431 Memslap testss multi\-get with both TCP and UDP. Because of
432 the different implementation of the ASCII protocol and binary protocol,
433 there are some differences between the two. For the ASCII protocol,
434 memaslap sends one “multi\-get” to the server once. For the
435 binary protocol, memaslap sends several single get commands
436 together as “multi\-get” to the server.
437 .SS UDP and TCP
438 .sp
439 Memslap testss both UDP and TCP. For TCP,
440 memaslap does not reconnect the memcached server if socket connections are
441 lost. If all the socket connections are lost or memcached server crashes,
442 memaslap will exit. If the user specifies the “\-\-reconnect”
443 option when socket connections are lost, it will reconnect them.
444 .sp
445 User can use “\-\-udp” to enable the UDP feature, but UDP comes with some
446 limitations:
447 .sp
448 UDP cannot set data more than 1400 bytes.
449 .sp
450 UDP is not testsed by the binary protocol because the binary protocol of
451 memcached does not tests that.
452 .sp
453 UDP doesn’t tests reconnection.
454 .SS Facebook test
455 .sp
456 Set data with TCP and multi\-get with UDP. Specify the following options:
457 .sp
458 "\-\-facebook \-\-division=50"
459 .sp
460 If you want to create thousands of TCP connections, specify the
461 .sp
462 "\-\-conn_sock=" option.
463 .sp
464 For example: \-\-facebook \-\-division=50 \-\-conn_sock=200
465 .sp
466 The above command means that memaslap will do facebook test,
467 each concurrency has 200 socket TCP connections and one UDP socket.
468 .sp
469 Memslap sets objects with the TCP socket, and multi\-gets 50
470 objects once with the UDP socket.
471 .sp
472 If you specify "\-\-division=50", the key size must be less that 25 bytes
473 because the UDP packet size is 1400 bytes.
474 .SS Replication test
475 .sp
476 For replication test, the user must specify at least two memcached servers.
477 The user can use “—rep_write=” option to enable feature.
478 .sp
479 For example:
480 .sp
481 \-\-servers=10.1.1.1:11211,10.1.1.2:11212 –rep_write=2
482 .sp
483 The above command means that there are 2 replication memcached servers,
484 memaslap will set objects to both server 0 and server 1, get
485 objects which are set to server 0 before from server 1, and also get objects
486 which are set to server 1 before from server 0. If server 0 crashes,
487 memaslap will only get objects from server 1. If server 0 comes
488 back to life again, memaslap will reconnect server 0. If both
489 server 0 and server 1 crash, memaslap will exit.
490 .SS Supports thousands of TCP connections
491 .sp
492 Start memaslap with "\-\-conn_sock=" or "\-n" to enable this
493 feature. Make sure that your system can tests opening thousands of files
494 and creating thousands of sockets. However, this feature does not tests
495 reconnection if sockets disconnect.
496 .sp
497 For example:
498 .sp
499 \-\-threads=8 \-\-concurrency=128 \-\-conn_sock=128
500 .sp
501 The above command means that memaslap starts up 8 threads, each
502 thread has 16 concurrencies, each concurrency has 128 TCP socket
503 connections, and the total number of TCP socket connections is 128 * 128 =
504 16384.
505 .SS Supports binary protocol
506 .sp
507 Start memaslap with "\-\-binary" or "\-B" options to enable this
508 feature. It testss all the above features except UDP, because the latest
509 memcached 1.3.3 does not implement binary UDP protocol.
510 .sp
511 For example:
512 .sp
513 \-\-binary
514 .sp
515 Since memcached 1.3.3 doesn\(aqt implement binary UDP protocol,
516 memaslap does not tests UDP. In addition, memcached 1.3.3 does not tests
517 multi\-get. If you specify "\-\-division=50" option, it just sends 50 get
518 commands together as “mulit\-get” to the server.
519 .SH CONFIGURATION FILE
520 .sp
521 This section describes the format of the configuration file. By default
522 when no configuration file is specified memaslap reads the default
523 one located at ~/.memaslap.cnf.
524 .sp
525 Below is a sample configuration file:
526 .sp
527 .nf
528 .ft C
529 \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
530 #comments should start with \(aq#\(aq
531 #key
532 #start_len end_len proportion
533 #
534 #key length range from start_len to end_len
535 #start_len must be equal to or greater than 16
536 #end_len must be equal to or less than 250
537 #start_len must be equal to or greater than end_len
538 #memaslap will generate keys according to the key range
539 #proportion: indicates keys generated from one range accounts for the total
540 generated keys
541 #
542 #example1: key range 16~100 accounts for 80%
543 # key range 101~200 accounts for 10%
544 # key range 201~250 accounts for 10%
545 # total should be 1 (0.8+0.1+0.1 = 1)
546 #
547 # 16 100 0.8
548 # 101 200 0.1
549 # 201 249 0.1
550 #
551 #example2: all keys length are 128 bytes
552 #
553 # 128 128 1
554 key
555 128 128 1
556 #value
557 #start_len end_len proportion
558 #
559 #value length range from start_len to end_len
560 #start_len must be equal to or greater than 1
561 #end_len must be equal to or less than 1M
562 #start_len must be equal to or greater than end_len
563 #memaslap will generate values according to the value range
564 #proportion: indicates values generated from one range accounts for the
565 total generated values
566 #
567 #example1: value range 1~1000 accounts for 80%
568 # value range 1001~10000 accounts for 10%
569 # value range 10001~100000 accounts for 10%
570 # total should be 1 (0.8+0.1+0.1 = 1)
571 #
572 # 1 1000 0.8
573 # 1001 10000 0.1
574 # 10001 100000 0.1
575 #
576 #example2: all value length are 128 bytes
577 #
578 # 128 128 1
579 value
580 2048 2048 1
581 #cmd
582 #cmd_type cmd_proportion
583 #
584 #currently memaslap only testss get and set command.
585 #
586 #cmd_type
587 #set 0
588 #get 1
589 #
590 #example: set command accounts for 50%
591 # get command accounts for 50%
592 # total should be 1 (0.5+0.5 = 1)
593 #
594 # cmd
595 # 0 0.5
596 # 1 0.5
597 cmd
598 0 0.1
599 1.0 0.9
600 .ft P
601 .fi
602 .SH FORMAT OF OUTPUT
603 .sp
604 At the beginning, memaslap displays some configuration information as follows:
605 .sp
606 servers : 127.0.0.1:11211
607 .sp
608 threads count: 1
609 .sp
610 concurrency: 16
611 .sp
612 run time: 20s
613 .sp
614 windows size: 10k
615 .sp
616 set proportion: set_prop=0.10
617 .sp
618 get proportion: get_prop=0.90
619 .SS Where
620 .sp
621 servers : "servers"
622 .INDENT 0.0
623 .INDENT 3.5
624 .sp
625 The servers used by memaslap.
626 .UNINDENT
627 .UNINDENT
628 .sp
629 threads count
630 .INDENT 0.0
631 .INDENT 3.5
632 .sp
633 The number of threads memaslap runs with.
634 .UNINDENT
635 .UNINDENT
636 .sp
637 concurrency
638 .INDENT 0.0
639 .INDENT 3.5
640 .sp
641 The number of concurrencies memaslap runs with.
642 .UNINDENT
643 .UNINDENT
644 .sp
645 run time
646 .INDENT 0.0
647 .INDENT 3.5
648 .sp
649 How long to run memaslap.
650 .UNINDENT
651 .UNINDENT
652 .sp
653 windows size
654 .INDENT 0.0
655 .INDENT 3.5
656 .sp
657 The task window size of each concurrency.
658 .UNINDENT
659 .UNINDENT
660 .sp
661 set proportion
662 .INDENT 0.0
663 .INDENT 3.5
664 .sp
665 The proportion of set command.
666 .UNINDENT
667 .UNINDENT
668 .sp
669 get proportion
670 .INDENT 0.0
671 .INDENT 3.5
672 .sp
673 The proportion of get command.
674 .UNINDENT
675 .UNINDENT
676 .sp
677 The output of dynamic statistics is something like this:
678 .sp
679 .nf
680 .ft C
681 \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
682 Get Statistics
683 Type Time(s) Ops TPS(ops/s) Net(M/s) Get_miss Min(us) Max(us)
684 Avg(us) Std_dev Geo_dist
685 Period 5 345826 69165 65.3 0 27 2198 203
686 95.43 177.29
687 Global 20 1257935 62896 71.8 0 26 3791 224
688 117.79 192.60
689
690
691 Set Statistics
692 Type Time(s) Ops TPS(ops/s) Net(M/s) Get_miss Min(us) Max(us)
693 Avg(us) Std_dev Geo_dist
694 Period 5 38425 7685 7.3 0 42 628 240
695 88.05 220.21
696 Global 20 139780 6989 8.0 0 37 3790 253
697 117.93 224.83
698
699
700 Total Statistics
701 Type Time(s) Ops TPS(ops/s) Net(M/s) Get_miss Min(us) Max(us)
702 Avg(us) Std_dev Geo_dist
703 Period 5 384252 76850 72.5 0 27 2198 207
704 94.72 181.18
705 Global 20 1397720 69886 79.7 0 26 3791 227
706 117.93 195.60
707 \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
708 .ft P
709 .fi
710 .SS Where
711 .sp
712 Get Statistics
713 .INDENT 0.0
714 .INDENT 3.5
715 .sp
716 Statistics information of get command
717 .UNINDENT
718 .UNINDENT
719 .sp
720 Set Statistics
721 .INDENT 0.0
722 .INDENT 3.5
723 .sp
724 Statistics information of set command
725 .UNINDENT
726 .UNINDENT
727 .sp
728 Total Statistics
729 .INDENT 0.0
730 .INDENT 3.5
731 .sp
732 Statistics information of both get and set command
733 .UNINDENT
734 .UNINDENT
735 .sp
736 Period
737 .INDENT 0.0
738 .INDENT 3.5
739 .sp
740 Result within a period
741 .UNINDENT
742 .UNINDENT
743 .sp
744 Global
745 .INDENT 0.0
746 .INDENT 3.5
747 .sp
748 Accumulated results
749 .UNINDENT
750 .UNINDENT
751 .sp
752 Ops
753 .INDENT 0.0
754 .INDENT 3.5
755 .sp
756 Total operations
757 .UNINDENT
758 .UNINDENT
759 .sp
760 TPS
761 .INDENT 0.0
762 .INDENT 3.5
763 .sp
764 Throughput, operations/second
765 .UNINDENT
766 .UNINDENT
767 .sp
768 Net
769 .INDENT 0.0
770 .INDENT 3.5
771 .sp
772 The rate of network
773 .UNINDENT
774 .UNINDENT
775 .sp
776 Get_miss
777 .INDENT 0.0
778 .INDENT 3.5
779 .sp
780 How many objects can’t be gotten
781 .UNINDENT
782 .UNINDENT
783 .sp
784 Min
785 .INDENT 0.0
786 .INDENT 3.5
787 .sp
788 The minimum response time
789 .UNINDENT
790 .UNINDENT
791 .sp
792 Max
793 .INDENT 0.0
794 .INDENT 3.5
795 .sp
796 The maximum response time
797 .UNINDENT
798 .UNINDENT
799 .sp
800 Avg:
801 .INDENT 0.0
802 .INDENT 3.5
803 .sp
804 The average response time
805 .UNINDENT
806 .UNINDENT
807 .sp
808 Std_dev
809 .INDENT 0.0
810 .INDENT 3.5
811 .sp
812 Standard deviation of response time
813 .UNINDENT
814 .UNINDENT
815 .sp
816 Geo_dist
817 .INDENT 0.0
818 .INDENT 3.5
819 .sp
820 Geometric distribution based on natural exponential function
821 .UNINDENT
822 .UNINDENT
823 .sp
824 At the end, memaslap will output something like this:
825 .sp
826 .nf
827 .ft C
828 \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
829 Get Statistics (1257956 events)
830 Min: 26
831 Max: 3791
832 Avg: 224
833 Geo: 192.60
834 Std: 116.23
835 Log2 Dist:
836 4: 0 10 84490 215345
837 8: 484890 459823 12543 824
838 12: 31
839
840 Set Statistics (139782 events)
841 Min: 37
842 Max: 3790
843 Avg: 253
844 Geo: 224.84
845 Std: 116.83
846 Log2 Dist:
847 4: 0 0 4200 16988
848 8: 50784 65574 2064 167
849 12: 5
850
851 Total Statistics (1397738 events)
852 Min: 26
853 Max: 3791
854 Avg: 227
855 Geo: 195.60
856 Std: 116.60
857 Log2 Dist:
858 4: 0 10 88690 232333
859 8: 535674 525397 14607 991
860 12: 36
861
862 cmd_get: 1257969
863 cmd_set: 139785
864 get_misses: 0
865 verify_misses: 0
866 verify_failed: 0
867 expired_get: 0
868 unexpired_unget: 0
869 written_bytes: 242516030
870 read_bytes: 1003702556
871 object_bytes: 152086080
872 packet_disorder: 0
873 packet_drop: 0
874 udp_timeout: 0
875
876 Run time: 20.0s Ops: 1397754 TPS: 69817 Net_rate: 59.4M/s
877 \-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-\-
878 .ft P
879 .fi
880 .SS Where
881 .sp
882 Get Statistics
883 .INDENT 0.0
884 .INDENT 3.5
885 .sp
886 Get statistics of response time
887 .UNINDENT
888 .UNINDENT
889 .sp
890 Set Statistics
891 .INDENT 0.0
892 .INDENT 3.5
893 .sp
894 Set statistics of response time
895 .UNINDENT
896 .UNINDENT
897 .sp
898 Total Statistics
899 .INDENT 0.0
900 .INDENT 3.5
901 .sp
902 Both get and set statistics of response time
903 .UNINDENT
904 .UNINDENT
905 .sp
906 Min
907 .INDENT 0.0
908 .INDENT 3.5
909 .sp
910 The accumulated and minimum response time
911 .UNINDENT
912 .UNINDENT
913 .sp
914 Max
915 .INDENT 0.0
916 .INDENT 3.5
917 .sp
918 The accumulated and maximum response time
919 .UNINDENT
920 .UNINDENT
921 .sp
922 Avg
923 .INDENT 0.0
924 .INDENT 3.5
925 .sp
926 The accumulated and average response time
927 .UNINDENT
928 .UNINDENT
929 .sp
930 Std
931 .INDENT 0.0
932 .INDENT 3.5
933 .sp
934 Standard deviation of response time
935 .UNINDENT
936 .UNINDENT
937 .sp
938 Log2 Dist
939 .INDENT 0.0
940 .INDENT 3.5
941 .sp
942 Geometric distribution based on logarithm 2
943 .UNINDENT
944 .UNINDENT
945 .sp
946 cmd_get
947 .INDENT 0.0
948 .INDENT 3.5
949 .sp
950 Total get commands done
951 .UNINDENT
952 .UNINDENT
953 .sp
954 cmd_set
955 .INDENT 0.0
956 .INDENT 3.5
957 .sp
958 Total set commands done
959 .UNINDENT
960 .UNINDENT
961 .sp
962 get_misses
963 .INDENT 0.0
964 .INDENT 3.5
965 .sp
966 How many objects can’t be gotten from server
967 .UNINDENT
968 .UNINDENT
969 .sp
970 verify_misses
971 .INDENT 0.0
972 .INDENT 3.5
973 .sp
974 How many objects need to verify but can’t get them
975 .UNINDENT
976 .UNINDENT
977 .sp
978 verify_failed
979 .INDENT 0.0
980 .INDENT 3.5
981 .sp
982 How many objects with insistent value
983 .UNINDENT
984 .UNINDENT
985 .sp
986 expired_get
987 .INDENT 0.0
988 .INDENT 3.5
989 .sp
990 How many objects are expired but we get them
991 .UNINDENT
992 .UNINDENT
993 .sp
994 unexpired_unget
995 .INDENT 0.0
996 .INDENT 3.5
997 .sp
998 How many objects are unexpired but we can’t get them
999 .UNINDENT
1000 .UNINDENT
1001 .sp
1002 written_bytes
1003 .INDENT 0.0
1004 .INDENT 3.5
1005 .sp
1006 Total written bytes
1007 .UNINDENT
1008 .UNINDENT
1009 .sp
1010 read_bytes
1011 .INDENT 0.0
1012 .INDENT 3.5
1013 .sp
1014 Total read bytes
1015 .UNINDENT
1016 .UNINDENT
1017 .sp
1018 object_bytes
1019 .INDENT 0.0
1020 .INDENT 3.5
1021 .sp
1022 Total object bytes
1023 .UNINDENT
1024 .UNINDENT
1025 .sp
1026 packet_disorder
1027 .INDENT 0.0
1028 .INDENT 3.5
1029 .sp
1030 How many UDP packages are disorder
1031 .UNINDENT
1032 .UNINDENT
1033 .sp
1034 packet_drop
1035 .INDENT 0.0
1036 .INDENT 3.5
1037 .sp
1038 How many UDP packages are lost
1039 .UNINDENT
1040 .UNINDENT
1041 .sp
1042 udp_timeout
1043 .INDENT 0.0
1044 .INDENT 3.5
1045 .sp
1046 How many times UDP time out happen
1047 .UNINDENT
1048 .UNINDENT
1049 .sp
1050 Run time
1051 .INDENT 0.0
1052 .INDENT 3.5
1053 .sp
1054 Total run time
1055 .UNINDENT
1056 .UNINDENT
1057 .sp
1058 Ops
1059 .INDENT 0.0
1060 .INDENT 3.5
1061 .sp
1062 Total operations
1063 .UNINDENT
1064 .UNINDENT
1065 .sp
1066 TPS
1067 .INDENT 0.0
1068 .INDENT 3.5
1069 .sp
1070 Throughput, operations/second
1071 .UNINDENT
1072 .UNINDENT
1073 .sp
1074 Net_rate
1075 .INDENT 0.0
1076 .INDENT 3.5
1077 .sp
1078 The average rate of network
1079 .UNINDENT
1080 .UNINDENT
1081 .SH OPTIONS
1082 .INDENT 0.0
1083 .TP
1084 .B \-s, \-\-servers=
1085 .
1086 List one or more servers to connect. Servers count must be less than
1087 threads count. e.g.: \-\-servers=localhost:1234,localhost:11211
1088 .TP
1089 .B \-T, \-\-threads=
1090 .
1091 Number of threads to startup, better equal to CPU numbers. Default 8.
1092 .TP
1093 .B \-c, \-\-concurrency=
1094 .
1095 Number of concurrency to simulate with load. Default 128.
1096 .TP
1097 .B \-n, \-\-conn_sock=
1098 .
1099 Number of TCP socks per concurrency. Default 1.
1100 .TP
1101 .B \-x, \-\-execute_number=
1102 .
1103 Number of operations(get and set) to execute for the
1104 given test. Default 1000000.
1105 .TP
1106 .B \-t, \-\-time=
1107 .
1108 How long the test to run, suffix: s\-seconds, m\-minutes, h\-hours,
1109 d\-days e.g.: \-\-time=2h.
1110 .TP
1111 .B \-F, \-\-cfg_cmd=
1112 .
1113 Load the configure file to get command,key and value distribution list.
1114 .TP
1115 .B \-w, \-\-win_size=
1116 .
1117 Task window size of each concurrency, suffix: K, M e.g.: \-\-win_size=10k.
1118 Default 10k.
1119 .TP
1120 .B \-X, \-\-fixed_size=
1121 .
1122 Fixed length of value.
1123 .TP
1124 .B \-v, \-\-verify=
1125 .
1126 The proportion of date verification, e.g.: \-\-verify=0.01
1127 .TP
1128 .B \-d, \-\-division=
1129 .
1130 Number of keys to multi\-get once. Default 1, means single get.
1131 .TP
1132 .B \-S, \-\-stat_freq=
1133 .
1134 Frequency of dumping statistic information. suffix: s\-seconds,
1135 m\-minutes, e.g.: \-\-resp_freq=10s.
1136 .TP
1137 .B \-e, \-\-exp_verify=
1138 .
1139 The proportion of objects with expire time, e.g.: \-\-exp_verify=0.01.
1140 Default no object with expire time
1141 .TP
1142 .B \-o, \-\-overwrite=
1143 .
1144 The proportion of objects need overwrite, e.g.: \-\-overwrite=0.01.
1145 Default never overwrite object.
1146 .UNINDENT
1147 .INDENT 0.0
1148 .TP
1149 .B \-R, \-\-reconnect
1150 .
1151 Reconnect tests, when connection is closed it will be reconnected.
1152 .TP
1153 .B \-U, \-\-udp
1154 .
1155 UDP tests, default memaslap uses TCP, TCP port and UDP port of
1156 server must be same.
1157 .TP
1158 .B \-a, \-\-facebook
1159 .
1160 Whether it enables facebook test feature, set with TCP and multi\-get with UDP.
1161 .TP
1162 .B \-B, \-\-binary
1163 .
1164 Whether it enables binary protocol. Default with ASCII protocol.
1165 .UNINDENT
1166 .INDENT 0.0
1167 .TP
1168 .B \-P, \-\-tps=
1169 .
1170 Expected throughput, suffix: K, e.g.: \-\-tps=10k.
1171 .TP
1172 .B \-p, \-\-rep_write=
1173 .
1174 The first nth servers can write data, e.g.: \-\-rep_write=2.
1175 .UNINDENT
1176 .INDENT 0.0
1177 .TP
1178 .B \-b, \-\-verbose
1179 .
1180 Whether it outputs detailed information when verification fails.
1181 .TP
1182 .B \-h, \-\-help
1183 .
1184 Display this message and then exit.
1185 .TP
1186 .B \-V, \-\-version
1187 .
1188 Display the version of the application and then exit.
1189 .UNINDENT
1190 .SH EXAMPLES
1191 .sp
1192 memaslap \-s 127.0.0.1:11211 \-S 5s
1193 .sp
1194 memaslap \-s 127.0.0.1:11211 \-t 2m \-v 0.2 \-e 0.05 \-b
1195 .sp
1196 memaslap \-s 127.0.0.1:11211 \-F config \-t 2m \-w 40k \-S 20s \-o 0.2
1197 .sp
1198 memaslap \-s 127.0.0.1:11211 \-F config \-t 2m \-T 4 \-c 128 \-d 20 \-P 40k
1199 .sp
1200 memaslap \-s 127.0.0.1:11211 \-F config \-t 2m \-d 50 \-a \-n 40
1201 .sp
1202 memaslap \-s 127.0.0.1:11211,127.0.0.1:11212 \-F config \-t 2m
1203 .sp
1204 memaslap \-s 127.0.0.1:11211,127.0.0.1:11212 \-F config \-t 2m \-p 2
1205 .SH HOME
1206 .sp
1207 To find out more information please check:
1208 \fI\%http://libmemcached.org/\fP
1209 .SH AUTHORS
1210 .sp
1211 Mingqiang Zhuang <\fI\%mingqiangzhuang@hengtiansoft.com\fP> (Schooner Technolgy)
1212 Brian Aker, <\fI\%brian@tangent.org\fP>
1213 .SH SEE ALSO
1214 .sp
1215 \fImemcached(1)\fP \fIlibmemcached(3)\fP
1216 .SH AUTHOR
1217 Brian Aker
1218 .SH COPYRIGHT
1219 2011, Brian Aker DataDifferential, http://datadifferential.com/
1220 .\" Generated by docutils manpage writer.
1221 .\"
1222 .