Redis QPS测试

 

1、计算qps：

1）redis发布版本中自带了redis-benchmark性能测试工具，可以使用它计算qps。示例：使用100个并发连接，发出100000个请求，每个请求的数据为2kb，测试host为127.0.0.1端口为6379的redis服务器性能：

/usr/bin/redis-benchmark -h 127.0.0.1 -p 6379 -c 100 -n 100000 -d 2

 

====== PING_INLINE ======
  100000 requests completed in 2.70 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

7.31% <= 1 milliseconds
32.79% <= 2 milliseconds
96.13% <= 3 milliseconds
99.85% <= 4 milliseconds
100.00% <= 4 milliseconds
36982.25 requests per second

====== PING_BULK ======
  100000 requests completed in 2.19 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

40.74% <= 1 milliseconds
53.82% <= 2 milliseconds
99.17% <= 3 milliseconds
99.96% <= 4 milliseconds
100.00% <= 5 milliseconds
100.00% <= 5 milliseconds
45599.63 requests per second

====== SET ======
  100000 requests completed in 2.61 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

27.14% <= 1 milliseconds
55.97% <= 2 milliseconds
99.05% <= 3 milliseconds
99.96% <= 4 milliseconds
100.00% <= 5 milliseconds
100.00% <= 5 milliseconds
38328.86 requests per second

====== GET ======
  100000 requests completed in 3.49 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

0.12% <= 1 milliseconds
45.66% <= 2 milliseconds
98.73% <= 3 milliseconds
99.87% <= 4 milliseconds
99.96% <= 5 milliseconds
99.97% <= 6 milliseconds
99.98% <= 7 milliseconds
100.00% <= 7 milliseconds
28645.09 requests per second

====== INCR ======
  100000 requests completed in 2.90 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

0.65% <= 1 milliseconds
9.45% <= 2 milliseconds
97.14% <= 3 milliseconds
99.71% <= 4 milliseconds
99.82% <= 5 milliseconds
99.89% <= 6 milliseconds
99.91% <= 7 milliseconds
99.98% <= 8 milliseconds
100.00% <= 8 milliseconds
34530.39 requests per second

====== LPUSH ======
  100000 requests completed in 2.36 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

22.02% <= 1 milliseconds
89.68% <= 2 milliseconds
99.92% <= 3 milliseconds
100.00% <= 4 milliseconds
42337.00 requests per second

====== RPUSH ======
  100000 requests completed in 3.10 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

43.54% <= 1 milliseconds
55.87% <= 2 milliseconds
99.31% <= 3 milliseconds
99.90% <= 4 milliseconds
99.95% <= 5 milliseconds
100.00% <= 5 milliseconds
32310.18 requests per second

====== LPOP ======
  100000 requests completed in 2.49 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

1.21% <= 1 milliseconds
61.49% <= 2 milliseconds
99.46% <= 3 milliseconds
100.00% <= 3 milliseconds
40128.41 requests per second

====== RPOP ======
  100000 requests completed in 2.73 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

24.75% <= 1 milliseconds
72.17% <= 2 milliseconds
99.57% <= 3 milliseconds
100.00% <= 3 milliseconds
36656.89 requests per second

====== SADD ======
  100000 requests completed in 3.45 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

5.84% <= 1 milliseconds
43.47% <= 2 milliseconds
97.75% <= 3 milliseconds
99.65% <= 4 milliseconds
99.87% <= 5 milliseconds
99.95% <= 6 milliseconds
99.99% <= 7 milliseconds
100.00% <= 8 milliseconds
100.00% <= 8 milliseconds
28960.32 requests per second

====== HSET ======
  100000 requests completed in 3.66 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

12.75% <= 1 milliseconds
32.22% <= 2 milliseconds
95.79% <= 3 milliseconds
99.40% <= 4 milliseconds
99.83% <= 5 milliseconds
99.96% <= 6 milliseconds
100.00% <= 6 milliseconds
27285.13 requests per second

====== SPOP ======
  100000 requests completed in 2.67 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

13.85% <= 1 milliseconds
35.88% <= 2 milliseconds
98.10% <= 3 milliseconds
99.73% <= 4 milliseconds
99.93% <= 5 milliseconds
99.98% <= 6 milliseconds
100.00% <= 7 milliseconds
100.00% <= 7 milliseconds
37425.15 requests per second

====== LPUSH (needed to benchmark LRANGE) ======
  100000 requests completed in 2.88 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

5.27% <= 1 milliseconds
49.51% <= 2 milliseconds
99.16% <= 3 milliseconds
99.82% <= 4 milliseconds
99.99% <= 5 milliseconds
100.00% <= 5 milliseconds
34686.09 requests per second

====== LRANGE_100 (first 100 elements) ======
  100000 requests completed in 2.58 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

30.03% <= 1 milliseconds
56.98% <= 2 milliseconds
99.46% <= 3 milliseconds
100.00% <= 3 milliseconds
38729.67 requests per second

====== LRANGE_300 (first 300 elements) ======
  100000 requests completed in 2.27 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

3.56% <= 1 milliseconds
99.29% <= 2 milliseconds
99.97% <= 3 milliseconds
100.00% <= 3 milliseconds
43994.72 requests per second

====== LRANGE_500 (first 450 elements) ======
  100000 requests completed in 2.00 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

61.48% <= 1 milliseconds
93.69% <= 2 milliseconds
99.69% <= 3 milliseconds
99.96% <= 4 milliseconds
100.00% <= 4 milliseconds
49975.02 requests per second

====== LRANGE_600 (first 600 elements) ======
  100000 requests completed in 2.64 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

2.13% <= 1 milliseconds
38.87% <= 2 milliseconds
98.63% <= 3 milliseconds
99.73% <= 4 milliseconds
99.79% <= 5 milliseconds
99.95% <= 6 milliseconds
99.99% <= 7 milliseconds
100.00% <= 7 milliseconds
37864.45 requests per second

====== MSET (10 keys) ======
  100000 requests completed in 2.79 seconds
  100 parallel clients
  2 bytes payload
  keep alive: 1

0.07% <= 1 milliseconds
57.01% <= 2 milliseconds
92.43% <= 3 milliseconds
99.84% <= 4 milliseconds
100.00% <= 5 milliseconds
35816.62 requests per second

我们关注结果最后一行：每秒35816.62个请求，既QPS4.4万；但这里的数据都只是测试数据，测出来的QPS不能代表实际生产的处理能力

 

2）测算redis处理实际生产请求的QPS/TPS

在实际生产中，我们需要关心这个指标，在我们的应用场景中，redis能够处理的最大的(QPS/TPS)是多少？测量redisQPS的方式有两种：

估计生产的报文大小，使用benchmark工具指定-d数据块大小来模拟；
使用redis-cli中info统计信息计算差值；redis-cli的info命令中有一项total_commands_processed表示：从启动到现在处理的所有命令总数，可以通过统计两次info指令间的差值来计算QPS：
 

watch -n 60'redis-cli -a MROhwkCyaGZbCrMTLAg2  info|grep total_commands_processe >> qps2.txt && date >>qps2.txt'

 

把每分钟的total_commands_processe记录到qps2.txt，打上时间戳得到数据两两相减，除以60秒。

 

注意：这个实时数据，存在波峰波谷，要采样几次，每次一段时间，这样才比较准。所以，一般使用下面命令

nohup watch -n 60'redis-cli -a MROhwkCyaGZbCrMTLAg2  info|grep total_commands_processe >> qps2.txt && date >>qps2.txt' > log 2>&1 &

 

2、内存使用情况：

redis是内存数据库，直接看info里的相关参数即可

used_memory:832784 # Redis 分配的内存总量 
used_memory_human:813.27K 
used_memory_rss:1896448 # Redis 分配的内存总量(包括内存碎片) 
used_memory_peak:832760  
used_memory_peak_human:813.24K #Redis所用内存的高峰值

 

3、redis连接数：

redis是内存数据库，直接看info里的相关参数即可

# Clients
connected_clients:28