VestaCP PHP-FPM VS Apache 09/11/2015

With the introduction of vestaCP 15, there is also a new installation wizard. It is now possible to set up a Nginx with PHP-fpm server. But why should we upgrade? The current setups still work? I tested a few setups. To see if an update is necessary. I set up two identical VPS systems, one using PHP-fpm another Apache.

Server info

Location: Amsterdam (NL)
CPU: 1
RAM: 765MB
OS: CentOS 7

• 

• 

Test 1: The standard HTML page.

Test 1 300 users in 1min (5 users/sec)

Apache response times: avg 86ms / max 544 ms / min 79 ms

PHP – FPM response times: avg 87 ms / max 467 ms / min 81 ms

Test 2 3000 users in 1 min (50 users/sec)

Apache response times: avg 83ms / max 456 ms / min 79 ms

PHP – FPM response times: avg 83ms / max 544 ms / min 80 ms

Test 3 10.000 users in 1 min (166 users/sec)

Apache response times: avg 82ms / max 373 ms / min 79 ms

PHP – FPM response times: avg 82ms / max 930 ms / min 79 ms

Test 4 20.000 users in 1 min (333 users/sec)

Apache response times: avg 166ms / max 1019 ms / min 157 ms

PHP – FPM response times: avg 165ms / max 514 ms / min 157 ms

From this test, we can’t conclude that much, only some variation in client/server. This is because we select a static page that was handled by Nginx.

Test 2: medium/large PHP/MySQL page

Test 1 300 users in 1min (5 users/sec)

Apache response times: avg 698 ms / max 1059 ms / min 407 ms

PHP – FPM response times: avg 588 ms / max 952 ms / min 454 ms

Test 2 3000 users in 1 min (50 users/sec)

This test failed due to a DOS.

Load Apache server:

Test 3 500 users in 1 min (8 users/sec)

Apache with Nginx cache response times: Avg 1082 ms / max 2871 ms / min 339 ms

PHP – FPM response times: Avg 942 ms / max 1856 ms / min 340 ms

Conclusion

We can clearly state that Nginx – PHP-FPM 10 to 20% faster than Apache, even if we use Nginx caching. The test site in phase 2 was maybe a little too heavy for this lightweight server with this traffic. During some tests, more than 5,000 queries were performed within one minute. The website is a copy of a news site (cyclingnieuws.be) white +10.000 news articles and +15.000 races with +250.000 results.

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

• 

Server upgrade:

After the execution of previous tests, I also wanted to know how this is going, on a slightly better server. I have upgraded the same servers to 8 vCPUs and 16GB of Ram.

Test 1: medium/large PHP/MySQL page

Test 1 300 users in 1min (5 users/sec)

Apache response times: Avg 373 ms / max 733 ms / min 326 ms

PHP – FPM response times: Avg 388 ms / max 791 ms / min 342 ms

Test 2 500 users in 1min (8 users/sec)

Apache response times: Avg 365 ms / max 643 ms / min 280 ms

PHP – FPM response times: Avg 373 ms / max 785 ms / min 260 ms

Test 3 1800 users in 1min (30 users/sec)

Apache response times: Avg 354 ms / max 845 ms / min 205 ms

PHP – FPM response times: Avg 298 ms / max 794 ms / min 193 ms

Test 4 3000 users in 1min (50 users/sec)

Apache response times: Avg 3605 ms / max 10084 ms / min 361 ms

PHP – FPM response times: Avg 338 ms / max 875 ms / min 259 ms

Test 5 6000 users in 1min (100 users/sec)

Apache response times: DOS

PHP – FPM response times: Avg 6448 ms / max 10238 ms / min 500 ms

Conclusion

In the beginning, we see a better working apache server, but because it only goes to a few milliseconds this is negligible. As of 30 users per second we see a difference in favor of PHP-fpm. As of 50 users per second apache can’t get his work done as opposed to PHP-FPM which can work to around 100 users per second.

• 

• 

• 

• 

• 

• 

• 

• 

•