New Benchmarks! LiteSpeed vs. Apache vs. nginx for Static Content

Our new static content benchmarks pit both versions of LiteSpeed Web Server against nginx and Apache, with LiteSpeed once again dusting the competition.

It’s the event everyone’s been waiting for: LiteSpeed Enterprise 4.2.5 vs. OpenLiteSpeed 1.2.6 vs. nginx 1.4.3 vs. Apache 2.4 vs. Apache 2.2. We tested these five web servers’ abilities to handle massive traffic requesting small static files (100 bytes) over both HTTP and HTTPS connections. The winner? LiteSpeed Enterprise.

Full results for HTTP and HTTPS.

Highlights

LSWS Enterprise showed the greatest gains with keep-alive connections enabled:

HTTP

245% faster than Apache 2.2 with pre-fork MPM
533% faster than Apache 2.4
67% faster than nginx

HTTPS

167% faster than Apache 2.2
302% faster than Apache 2.4
47% faster than nginx

LiteSpeed Enterprise also tested a little faster than OpenLiteSpeed in all tests, though OpenLiteSpeed still bested all other competitors.

Discussion

We believe the test shows two things:

A stark difference between event-driven and process-based web servers: Both LiteSpeed and nginx severely out-performed Apache, even with event MPM.
Not all event-driven web servers are made the same: nginx just does not have the speed of LiteSpeed’s highly optimized code.

In the test we tried to remove as many other variables as we could, using the same hardware and purposely keeping the file size tiny. The goal was to demonstrate pure speed differences when serving your static content. This is, of course, just one facet of web service and, in future tests, we will examine other situations and types of content, including bigger files and dynamic pages.

Reactions, Please

This is the first in a series of tests that will examine performance with many kinds of content. We hope this test will spark conversation. Let us know your reactions on our forum. You can find the test configurations and hardware specs with the benchmark results. We’d be happy to hear any objections you have or feedback on the configurations. We encourage you to reproduce the test yourself or suggest any changes you would like to see.

Tags:

Categories:Benchmarks , LiteSpeed Web Server , OpenLiteSpeed

Comments

Hayden James January 31st, 2014
This is why I don’t go by benchmarks especially when all the details of hardware, software, configuration, disclosure of what “static content”. I agree misconfigured.
- Michael February 3rd, 2014
  Hi Hayden,
  All the hardware specs are available on the site. The configurations can also be downloaded there. If you think the software was misconfigured, we’re happy to respond to any specific objections you have as to the software’s configs.
  Cheers,
  Michael
Bill Stevens January 7th, 2014
You’re telling me that prefork is faster than event? I think your event-mpm Apache instance was misconfigured
- Michael January 8th, 2014
  Hi Bill,
  That is what the results of our benchmark suggest. We were a little surprised, too. It should be noted, though, that results may vary when used in the (much more complicated) real world.
  The configs we used are available for download. If anyone has any suggestions as to how they can be improved (not to optimize one instance in comparison to another, but rather to make sure all instances are competing on even ground), we are very interested.
  Cheers,
  Michael
Sadek November 27th, 2013
Planning to install litespeed WHM extention in my VPS server.
Let’s see what happens.
- Michael November 27th, 2013
  Be sure to let us know! We have a feeling you’ll be pleasantly surprised.