OpenLitespeed vs Nginx performance under heavy load testing

Misao

New Member
#1
What’s your experience with OpenLitespeed/Litespeed vs Nginx lately? I’ve just made some load tests comparing both of them and the results are unexpected.

(Using 5$ Vultr droplet Singapore location. Default wordpress 5.3 install. Php 7.3)

I’m using Gridpane for Nginx stack and Cyberpanel for Openlitespeed stack.

Loader.io test

1st test 250 clients per sec: OpenLitespeed did pretty good with only 261ms average response time.

Nginx 343ms average response time.

2nd test 500 clients per sec: OpenLitespeed fall hard with 4992ms avg response time and 42% errors.

Nginx was very stable with 367ms avg respone time and zero error.

3rd test 1000 clients per sec: OpenLitespeed couldn’t finnish the test and crashed. I have to reatart it few time but the result was the same. Switched to Enterprise version and 4529ms avg response time with 42% errors.

Nginx still did good with 2046ms avg response time with only 2% errors.

I used to believe that there is no way nginx can beat OpenLitespeed/Litespeed in load test. Or did I miss something? Do you guys have any suggestion for me?

The reason why I choose Cyberpanel and Gridpane? Because they are easy to install for normal user like me.

Updated: Just upgrade VPS to 2vCPU 4GB Ram, running Litespeed Enterprise with LS Cache enabled. Results:
2nd test with 500 clients per sec: 1366 ms avg resp 10.5% err rate.
3rd test with 1000 clients per sec: 4610 ms avg resp 43.6% err rate.

A 20$ VPS running Cyberpanel with Litespeed Enterprise still can't beat 5$ Nginx VPS with Gridpane. Is it true or there is something wrong with Cyberpanel and OpenLitespeed?
 

Cold-Egg

Administrator
#2
Hi,

Default "Max Connections" and "Max SSL connections" values are still low, e.g. 2k on CyberPanel if I remember correctly. To do the benchmark with it, please increase both to like 10k and then try it again.

Let us know if it helps.

Best
 

Misao

New Member
#3
Hi,

Cyberpanel's default values for "Max Connections: 20000" and "Max SSL Connections: 10000".

Just try to spin a new Vultr HF droplet which has 2vCPU (3.7Ghz) and 4GB Ram, Los Angeles location and it can hold up to 2000 clients per sec test. More than that, OLS crashed.

But a good thing about OLS is: It was very stable until it crashed. Avg response time was the same (59-65ms) from 50 to 2000 clients per sec tests.

I always like OLS so is there any suggestion or config that I can do to make it on par with Nginx (Gridpane stack)?

Thanks.
 
Top