redis-benchmark压力测试

redis-benchmark压力测试

redis-benchmark是redis官方提供的压测工具，安装好redis后，默认安装。使用简便。

语法：

Usage: redis-benchmark [-h <host>] [-p <port>] [-c <clients>] [-n <requests]> [-k <boolean>]

模拟20个客户端，100000次请求

redis-benchmark -h 192.168.1.1 -p 6379 -n 100000 -c 20

模拟1000000次请求，生成100000000个set结构

redis-benchmark -t set -n 1000000 -r 100000000

模拟ping，set，get 各100000次，结果输出到csv文件

redis-benchmark -t ping,set,get -n 100000 --csv

模拟100000次键foo的存储性能

redis-benchmark -n 100000 -q script load "redis.call('set','foo','bar')"

模拟一下十万次请求：
redis-benchmark -n 100000 -q

模拟一下百万级访问近百万key：
[root@vm-ArthurGuo-1 ~]# redis-benchmark -n 1000000 -r1000000 -q

模拟一个万级用户的并发：
 redis-benchmark -c 10000 -n 1000000 -r 1000000 -q

redis-benchmark --help

Usage: redis-benchmark [-h <host>] [-p <port>] [-c <clients>] [-n <requests]> [-k <boolean>]

 -h <hostname>      Server hostname (default 127.0.0.1)  --主机ip地址

 -p <port>          Server port (default 6379) --端口

 -s <socket>        Server socket (overrides host and port) --socket（如果测试在服务器上测可以用socket方式）

 -c <clients>       Number of parallel connections (default 50) --客户端连接数

 -n <requests>      Total number of requests (default 10000) --总请求数

 -d <size>          Data size of SET/GET value in bytes (default 2) --set、get的value大小

 -dbnum <db>        SELECT the specified db number (default 0) --选择哪个数据库测试（一般0-15）

 -k <boolean>       1=keep alive 0=reconnect (default 1) --是否采用keep alive模式

 -r <keyspacelen>   Use random keys for SET/GET/INCR, random values for SADD --随机产生键值时的随机数范围

  Using this option the benchmark will expand the string __rand_int__

  inside an argument with a 12 digits number in the specified range

  from 0 to keyspacelen-1. The substitution changes every time a command

  is executed. Default tests use this to hit random keys in the

  specified range.

 -P <numreq>        Pipeline <numreq> requests. Default 1 (no pipeline). --pipeline的个数（如果使用pipeline会把多个命令封装在一起提高效率）

 -q                 Quiet. Just show query/sec values --仅仅查看每秒的查询数

 --csv              Output in CSV format --用csv方式输出

 -l                 Loop. Run the tests forever --循环次数

 -t <tests>         Only run the comma separated list of tests. The test --指定命令

                    names are the same as the ones produced as output.

 -I                 Idle mode. Just open N idle connections and wait. --仅打开n个空闲链接

Examples:

 Run the benchmark with the default configuration against 127.0.0.1:6379:

   $ redis-benchmark

 Use 20 parallel clients, for a total of 100k requests, against 192.168.1.1:

   $ redis-benchmark -h 192.168.1.1 -p 6379 -n 100000 -c 20  --测试set、get、mset、sadd等场景下的性能

 Fill 127.0.0.1:6379 with about 1 million keys only using the SET test:

   $ redis-benchmark -t set -n 1000000 -r 100000000  --测试set随机数的性能

 Benchmark 127.0.0.1:6379 for a few commands producing CSV output:

   $ redis-benchmark -t ping,set,get -n 100000 --csv  --使用csv的输出方式测试

 Benchmark a specific command line:

   $ redis-benchmark -r 10000 -n 10000 eval 'return redis.call("ping")' 0 --测试基本命令的速度

 Fill a list with 10000 random elements:

   $ redis-benchmark -r 10000 -n 10000 lpush mylist __rand_int__  --测试list入队的速度

 On user specified command lines __rand_int__ is replaced with a random integer

 with a range of values selected by the -r option.

下面我就测下我的笔记本电脑的redis性能：

[root@db1 ~]# redis-benchmark -h 127.0.0.1 -p 6379 -n 100000 -c 20

====== PING_INLINE ======

  100000 requests completed in 1.09 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.86% <= 1 milliseconds

100.00% <= 2 milliseconds

100.00% <= 2 milliseconds

91659.03 requests per second

====== PING_BULK ======

  100000 requests completed in 1.07 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.94% <= 1 milliseconds

100.00% <= 1 milliseconds

93545.37 requests per second

====== SET ======

  100000 requests completed in 1.03 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.78% <= 1 milliseconds

100.00% <= 1 milliseconds

97087.38 requests per second

====== GET ======

  100000 requests completed in 1.10 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.81% <= 1 milliseconds

100.00% <= 1 milliseconds

90909.09 requests per second

====== INCR ======

  100000 requests completed in 1.09 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.86% <= 1 milliseconds

100.00% <= 1 milliseconds

91911.76 requests per second

====== LPUSH ======

  100000 requests completed in 1.07 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.85% <= 1 milliseconds

100.00% <= 1 milliseconds

93808.63 requests per second

====== LPOP ======

  100000 requests completed in 1.01 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.89% <= 1 milliseconds

100.00% <= 1 milliseconds

98522.17 requests per second

====== SADD ======

  100000 requests completed in 1.04 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.76% <= 1 milliseconds

100.00% <= 1 milliseconds

96153.85 requests per second

====== SPOP ======

  100000 requests completed in 1.11 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.92% <= 1 milliseconds

100.00% <= 1 milliseconds

90171.33 requests per second

====== LPUSH (needed to benchmark LRANGE) ======

  100000 requests completed in 1.09 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.82% <= 1 milliseconds

100.00% <= 1 milliseconds

92081.03 requests per second

====== LRANGE_100 (first 100 elements) ======

  100000 requests completed in 2.53 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.91% <= 1 milliseconds

100.00% <= 2 milliseconds

100.00% <= 2 milliseconds

39603.96 requests per second

====== LRANGE_300 (first 300 elements) ======

  100000 requests completed in 5.17 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

91.01% <= 1 milliseconds

99.94% <= 2 milliseconds

100.00% <= 2 milliseconds

19346.10 requests per second

====== LRANGE_500 (first 450 elements) ======

  100000 requests completed in 7.41 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

61.54% <= 1 milliseconds

98.36% <= 2 milliseconds

99.96% <= 3 milliseconds

100.00% <= 4 milliseconds

100.00% <= 4 milliseconds

13498.92 requests per second

====== LRANGE_600 (first 600 elements) ======

  100000 requests completed in 9.49 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

41.24% <= 1 milliseconds

91.89% <= 2 milliseconds

99.78% <= 3 milliseconds

100.00% <= 4 milliseconds

100.00% <= 4 milliseconds

10541.85 requests per second

====== MSET (10 keys) ======

  100000 requests completed in 1.68 seconds

  20 parallel clients

  3 bytes payload

  keep alive: 1

99.28% <= 1 milliseconds

100.00% <= 1 milliseconds

59382.42 requests per second

从以上可以看出，20个客户端，每种场景均有100000次请求：ping、set、get、lpush、lpop、spop等都达到90000多rps，但lrange前100、300、500等就比较慢了，才10000多rps。

再测下set的速度：

[root@db1 ~]# redis-benchmark -t set -n 1000000 -r 100000000

====== SET ======

  1000000 requests completed in 10.56 seconds

  50 parallel clients

  3 bytes payload

  keep alive: 1

98.65% <= 1 milliseconds

99.90% <= 2 milliseconds

99.99% <= 3 milliseconds

100.00% <= 3 milliseconds

94741.83 requests per second

每秒94741次，非常快

再来测试下list的入队速度：

[root@db1 ~]# redis-benchmark -r 100000 -n 100000 lpush mylist __rand_int__

====== lpush mylist __rand_int__ ======

  100000 requests completed in 0.97 seconds

  50 parallel clients

  3 bytes payload

  keep alive: 1

98.83% <= 1 milliseconds

100.00% <= 1 milliseconds

102774.92 requests per second

超过了10w次。