laravel-fpm、laravel-octane、webman性能对比

性能讨论#

我之前认为 laravel-octane 性能和 webman 应该在同一个数量级，没想到差距这么大
大家怎么看呢

说明#

配置: 12 核 8G
运行环境: Docker
测试方式：请求直接返回 Json
PHP 版本: 8.2 (未开启 opcache)
Laravel 版本: 10.42.0 (未做任何优化)
测试方法：请求直接返回 hello world

laravel-fpm#

附上 php-fpm 的配置

pm = dynamic
pm.max_children = 200
pm.start_servers = 50
pm.min_spare_servers = 20
pm.max_spare_servers = 200

ab -n 10000 -c 1000 测试结果如下

ab -n 10000 -c 1000 http://localhost:80/hello
This is ApacheBench, Version 2.3 <$Revision: 1903618 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking localhost (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
^@Completed 9000 requests
Completed 10000 requests
Finished 10000 requests


Server Software:        nginx/1.25.3
Server Hostname:        localhost
Server Port:            80

Document Path:          /hello
Document Length:        33 bytes

Concurrency Level:      1000
Time taken for tests:   80.732 seconds
Complete requests:      10000
Failed requests:        0
Total transferred:      12210000 bytes
HTML transferred:       330000 bytes
Requests per second:    123.87 [#/sec] (mean)
Time per request:       8073.239 [ms] (mean)
Time per request:       8.073 [ms] (mean, across all concurrent requests)
Transfer rate:          147.70 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    2   6.4      0      28
Processing:    84 7735 1590.4   7859   17599
Waiting:       56 7735 1590.4   7859   17599
Total:         84 7737 1586.4   7859   17599

Percentage of the requests served within a certain time (ms)
  50%   7859
  66%   8002
  75%   8107
  80%   8182
  90%   8419
  95%   8709
  98%   9222
  99%  15093
 100%  17599 (longest request)

laravel-octane#

ab -n 10000 -c 1000 测试结果如下

ab -n 10000 -c 1000 http://localhost:8000/hello
This is ApacheBench, Version 2.3 <$Revision: 1903618 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking localhost (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Completed 10000 requests
Finished 10000 requests


Server Software:
Server Hostname:        localhost
Server Port:            8000

Document Path:          /hello
Document Length:        33 bytes

Concurrency Level:      1000
Time taken for tests:   26.674 seconds
Complete requests:      10000
Failed requests:        0
Total transferred:      10540000 bytes
HTML transferred:       330000 bytes
Requests per second:    374.90 [#/sec] (mean)
Time per request:       2667.369 [ms] (mean)
Time per request:       2.667 [ms] (mean, across all concurrent requests)
Transfer rate:          385.88 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    2   6.3      0      28
Processing:    12 2519 610.6   2651    3657
Waiting:        3 2517 609.7   2651    3657
Total:         31 2521 604.6   2651    3657

Percentage of the requests served within a certain time (ms)
  50%   2651
  66%   2657
  75%   2661
  80%   2665
  90%   3082
  95%   3119
  98%   3134
  99%   3140
 100%   3657 (longest request)

webman#

ab -n 10000 -c 1000 测试结果如下

ab -n 10000 -c 1000 http://localhost:8001/hello
This is ApacheBench, Version 2.3 <$Revision: 1903618 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking localhost (be patient)
Completed 1000 requests
Completed 2000 requests
Completed 3000 requests
Completed 4000 requests
Completed 5000 requests
Completed 6000 requests
Completed 7000 requests
Completed 8000 requests
Completed 9000 requests
Completed 10000 requests
Finished 10000 requests


Server Software:        workerman
Server Hostname:        localhost
Server Port:            8001

Document Path:          /hello
Document Length:        33 bytes

Concurrency Level:      1000
Time taken for tests:   0.625 seconds
Complete requests:      10000
Failed requests:        0
Total transferred:      1470000 bytes
HTML transferred:       330000 bytes
Requests per second:    16009.27 [#/sec] (mean)
Time per request:       62.464 [ms] (mean)
Time per request:       0.062 [ms] (mean, across all concurrent requests)
Transfer rate:          2298.21 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0   24   7.1     25      39
Processing:     7   35  23.0     29     140
Waiting:        1   27  23.6     20     126
Total:         29   59  18.9     55     155

Percentage of the requests served within a certain time (ms)
  50%     55
  66%     56
  75%     57
  80%     58
  90%    101
  95%    113
  98%    118
  99%    124
 100%    155 (longest request)

laravel laravel-octane webman

讨论数量: 22

JaguarJack

Laravel 9.x 译者 940 声望 / 开发 @ 家里蹲开发公司

这是干嘛的这么低？octane

1年前评论

skarner （楼主）

这也是大大出乎我意料的地方，octane 和 webman 相差太大了

@skarner 不过有一点，laravel-fpm 请求过去CPU直接打满了，但是octane CPU占用率相对来说比较低（印象中大概50%-60%的样子，忘了记录）

GDDD

110 声望 / 工头 @ 工地

webman 用的人都说真香，你还在这压测对比

我想知道把 Laravel 优化到极致，和webman的性能差距到底有多大

fatrbaby

274 声望 / 首席换水官 @ 名剑风流

webman 核心的代码质量如果能达到 Laravel 的水准，甚至达到 Symfony 的水准，我立马跪下来叫爸爸

那不可能，差远了

mengmeng_phper

内核代码我看了，写的跟laravel差的有点远

论实现框架代码的优雅，Webman 确实远不及 Laravel ，但是他快得离谱啊

如此甚好

98 声望

咋感觉不太相信，你 octane 用的哪个，swoole 还是

octane 用了 RoadRunner 和 Swoole 都试过，这两个数据都非常接近，所以没有做区分

不负岁月

57 声望

那么，该怎么选！

之前我考虑的是，Octane应该能非常大的弥补 Laravel 重和慢的缺点。实际测试发现提升和我预期的有较大差距。可能是我对Octane的预期太高了

不负岁月（作者）

@skarner github.com/joanhey/AdapterMan 可以试试这个

尝试了 AdapterMan，发现用AB 测试会报错： apr_pollset_poll: The timeout specified has expired (70007)

但是直接CURL请求返回是正常的，我尝试去找原因，发现weman的作者说 “wls2本身性能很差，尤其是有磁盘读写的时候。因为wls2成为瓶颈，压测结果相差不是不大。”

我就是用的WSL，可能这个结果不具有太大的参考意义

晏南风

见习助教 149 声望

webman 和 laravel octane 进程数分别多少

webman 进程数是默认的，CPU数量 * 4
octane 进程数量是按文档的CPU数量
我尝试了把 octane 的进程数也该为CPU*4 ，性能和之前没什么明显差别

33qis

23 声望

第一： ab 测我感觉不准

第二：fpm 模式配置可以改成 static

第三：opcache 以及 laravel 优化都要开，毕竟除了 fpm 其他都是内存型的

第四：可以考虑 io 阻塞问题

测试了一下，把php-fpm 改成了 static, laravel 除opcache以外的优化都开了，提升非常有限。 Requests per second从123提升到了130，开启opcache可能会有较大提升，晚点儿我再测试一下

罐装仙人掌CuratorC

Laravel 9.x 译者 81 声望 / phper ->gopher @ 罐罐头

可以再试试 hyperf

Larva

这个测试没意义，要考虑框架初始化时候都干了什么，只输出 hello world 那就需要做优化

xini2603

54 声望

webman 的确快，不过最终差距没有这么大，最大的优点是 webman 占用负载情况远完小于 fpm fpm 压测负载就 100 了，webman 50% 左右，另外我现在的同一套代码同时运行在 webman 与 laravel. 在 webman 引用了 laravel 的 orm,cache,auth, 在 laravel 下面建个目录放 webman，webman 读取 laravel 的 env 配置，并运行 laravel 的业务部分代码，以实现只须要在 laravel 下面写代码，同时可在 webman 中运行，对须要高并发的就通过 ngixn 把业务转发到 webman 下面，

讨论应以学习和精进为目的。请勿发布不友善或者负能量的内容，与人为善，比聪明更重要！

帮助

laravel-fpm、laravel-octane、webman性能对比

性能讨论#

说明#

laravel-fpm#

laravel-octane#

webman#

推荐文章：

社区赞助商

关于 LearnKu

资源推荐

服务提供商

其他信息

laravel-fpm、laravel-octane、webman性能对比

性能讨论#

说明#

laravel-fpm#

laravel-octane#

webman#

推荐文章：

社区赞助商

关于 LearnKu

资源推荐

服务提供商

其他信息

请登录