Updated benchmarks ?

wanah

Well-Known Member
#1
Hello,

Any chance the benchmarks on the litespeed website could be updated ?

http://www.litespeedtech.com/web-server-performance-comparison-litespeed-2.1-vs.html

These benchmarks are very old.

I would be interested to see how Litespeed 4.2 comparess to Apache 2.4 and Nginx 1.4 or 1.5, I'm sure Litespeed has made quite a leap forward since version 2…

I think every new major version of Litespeed shoud get a benchmark and any version that improves either PHP's speeds or static files speeds :)
 

Michael

Well-Known Member
Staff member
#2
Hi Wanah,

We definitely agree. Those are far too out of date. We'll be getting on some new benchmarks soon, especially with our new Python WSGI support.

We don't do benchmarks too often because they don't prove very much — it's incredibly hard to set up a benchmark that reflects the real world, so there will always be arguments over what's faster in what situation. That said, we could all use more information.

As I said, we'll be doing some new benchmarks for dynamic content, but we always appreciate any benchmarks anyone else performs. If anyone's got a benchmark, we'd love to share it.

Michael
 
#3
Why not setup a test server with a Wordpress install and cache database queries in memcache as much as possible, then benchmark the time it takes to load a sample blogpost?

While database queries can cause variance in the benchmark, however it would provide a more real world performance benchmark than the simple hello world php pages.
 

wanah

Well-Known Member
#4
I think there are multiple situations that should be benched :

1) HTML page with almost no contents
2) Large image
3) Simple PHP script with just : <?php echo "hello world"?>
4) Complicated PHP script without any database access

With a decent cache plugin Wordpress sends all information as it it was a plain html page.

Also some scripts run much faster on recent versions of PHP. Recently I tested a script that used 15MB on PHP5.3, 10MB on PHP 5.4 and 5MB on PHP 5.5 and took 9ms on PHP 5.3, 6ms on PHP 5.4 and 3ms on PHP 5.5.

So this isn't all about the webservers API, the most important thing is how quickly the API passes on the data to the browser.
 

Michael

Well-Known Member
Staff member
#5
Why not setup a test server with a Wordpress install and cache database queries in memcache as much as possible, then benchmark the time it takes to load a sample blogpost?

While database queries can cause variance in the benchmark, however it would provide a more real world performance benchmark than the simple hello world php pages.
Hi Yogesh,

Yes, adding WordPress into the mix would be a step closer to a real world situation, but it wouldn't be much closer. It would only add one of the multitude of factors that slow down sites. It would not really counter the argument that the benchmarks don't reflect the real world -- and it would make less clear the differences, if any, in the results.

To show real world differences we are currently recruiting some sites for case studies that will show the benefits of switching to LiteSpeed on a production site. Benchmarks, on the other hand, should be a "pure" comparison of two or more softwares, and thus it's important to keep them simple.

michael
 
Last edited:

eva2000

Well-Known Member
#6
Would be great if LSWS/OLS admin console had inbuilt server usage stats charts and history for cpu load, memory usage and disk usage. Would be handy for benchmarks or long term trend analysis/benchmark type case studies :D
 
Last edited:

eva2000

Well-Known Member
#8
Only used cacti and munin myself. But cacti and nagios have good support for 3rd party templates i.e. Percona's monitoring tool plugins http://www.percona.com/doc/percona-monitoring-plugins/

cacti plugins

  • Percona Monitoring Plugins for Cacti
  • Frequently Asked Questions on Cacti Templates
  • Installing Percona Monitoring Plugins for Cacti
  • Customizing Percona Monitoring Plugins for Cacti
  • Upgrading Percona Monitoring Plugins for Cacti
  • Percona MySQL Monitoring Template for Cacti
  • Percona Galera/MySQL Monitoring Template for Cacti
  • Installing SSH-Based Templates
  • Percona Apache Monitoring Template for Cacti
  • Percona JMX Monitoring Template for Cacti
  • Percona Linux Monitoring Template for Cacti
  • Percona Memcached Monitoring Template for Cacti
  • Percona MongoDB Monitoring Template for Cacti
  • Percona Nginx Monitoring Template for Cacti
  • Percona OpenVZ Monitoring Template for Cacti
  • Percona Redis Monitoring Template for Cacti
  • Cacti Templates Developer Documentation

Might was well import and use percona's plugins so you have one less thing to maintain. And modify Apache template for LiteSpeed/OLS :D
 
Last edited:

Michael

Well-Known Member
Staff member
#9
Hey George,

Those sound like great things to have, especially for benchmarking/case studies, but I don't think they're part of what the web server should provide. There are lots of applications and many control panels already have plugins with that functionality.

Still, I, for one, would be really, really, really interested in benchmarks showing memory usage and CPU load over time with different servers. Really interested. That, I think, would go to the heart of what makes LSWS great.

m
 

Michael

Well-Known Member
Staff member
#11
The Cacti template we have now doesn't work? I know you posted on the Google Group about this before, but I'm a little foggy on what you need updated. Is there something about the wiki that doesn't work for CentOS 6? (Sorry I haven't had time to test it myself.)

m
 
Top