我们都知道http1.1是支持keepalive，这样就可以减少建连的次数，可以提高性能。
于是简单在单机上做了一个测试。
我们先使用一个没有keepalive的配置。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
| upstream test_server {
server 127.0.0.1:8999;
server 127.0.0.1:8997;
keepalive 32;
}
server {
listen 8998;
server_name _;
access_log /opt/server/nginx/log/8998-access.log main;
error_log /opt/server/nginx/log/8998-error.log;
location / {
proxy_pass http://test_server;
}
}
server {
listen 8999;
server_name _;
access_log /opt/server/nginx/log/8999-access.log main;
error_log /opt/server/nginx/log/8999-error.log;
location / {
return 200 '{"status":"OK","entities":[]}';
}
}
server {
listen 8997;
server_name _;
access_log /opt/server/nginx/log/8997-access.log main;
error_log /opt/server/nginx/log/8997-error.log;
location / {
return 200 '{"status":"OK","entities":[]}';
}
}
|
测试结果如下：
1
2
3
4
5
6
7
8
9
| $ wrk -t1 -c32 -d60s http://127.0.0.1:8998/
Running 1m test @ http://127.0.0.1:8998/
1 threads and 32 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 1.61ms 2.14ms 76.87ms 99.52%
Req/Sec 21.31k 2.22k 26.73k 69.00%
1272221 requests in 1.00m, 260.85MB read
Requests/sec: 21202.84
Transfer/sec: 4.35MB
|
我们再打开keepalive的设置
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
| upstream test_server {
server 127.0.0.1:8999;
server 127.0.0.1:8997;
keepalive 32;
}
server {
listen 8998;
server_name _;
access_log /opt/server/nginx/log/8998-access.log main;
error_log /opt/server/nginx/log/8998-error.log;
location / {
proxy_pass http://test_server;
proxy_http_version 1.1;
proxy_set_header Connection "";
}
}
server {
listen 8999;
server_name _;
access_log /opt/server/nginx/log/8999-access.log main;
error_log /opt/server/nginx/log/8999-error.log;
location / {
return 200 '{"status":"OK","entities":[]}';
}
}
server {
listen 8997;
server_name _;
access_log /opt/server/nginx/log/8997-access.log main;
error_log /opt/server/nginx/log/8997-error.log;
location / {
return 200 '{"status":"OK","entities":[]}';
}
}
|
这次测试结果如下：
1
2
3
4
5
6
7
8
9
| $ wrk -t1 -c32 -d60s http://127.0.0.1:8998/
Running 1m test @ http://127.0.0.1:8998/
1 threads and 32 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 817.00us 3.10ms 107.27ms 99.47%
Req/Sec 49.73k 6.16k 74.11k 75.50%
2968158 requests in 1.00m, 608.58MB read
Requests/sec: 49468.57
Transfer/sec: 10.14MB
|
我们可以看到一个测试结果是每秒2万次，一个是每秒49000次。差了大概2.5倍左右了。
原因我们应该很容易明白，一个是有很多的时间花在了3次握手和4次挥手上了。
虽然在localhost上这些时间都非常短，但是一旦实际中花的时间是不会少的。
而keepalive的我们就可以很明显的看到一个tcp连接进行了多少次的get请求
数了一下是100次GET请求，这个估计是因为我是nginx 1.16.1的缘故，在1.19.1以后是可以单独设置的。keepalive_time时长也可以单独来进行设置。