Update: Checkout the first & only cPanel Varnish plugin: http://www.unixy.net/varnish
This page is part of our Accelerate Your Server campaign. Be sure to read all about it here:
http://www.unixy.net/accelerate-your-server
Everyone knows that the Apache Web server is lagging behind
Litespeed in terms of responsiveness
and performance. Not by much though! But what you probably do not know is that we have built a
platform around Apache that runs circles around Litespeed,
95% faster to be exact. In this page,
we are going to share benchmark results and technology we are using to help save our customers
a few bucks but most importantly outperform litespeed. Before we take a look at the numbers,
it is worth mentioning the advantages of running Apache over Litespeed:
Apache is free and open source. Litespeed is closed source.
Apache has a strong online user and developer community.
Apache costs zip! Litespeed goes for $699/yr for a 4-CPU license.
Apache has ample and abundant documentation.
The Apache Foundation is a not-for-profit corporation.
For our benchmark hosting, we picked a dual-core Opteron server with 2GB RAM from
our
Bargain Bin and created two VMs. Each VM has 1GB RAM assigned and enough disk
space. One VM contains Litespeed Enterprise + PHP LSAPI. The other VM contains
Red Hat Apache from RPM and some other software, which we shall cover shortly.
Apache Config
Apache/2.2.3 x86 64-bit w/ Varnish
PHP 5 DSO
Prefork MPM
MaxClients 300
Virtuozzo VM
Litespeed Config
LiteSpeed/4.0.15 Enterprise x86 64-bit
PHP 5 LSAPI
Max Connections 2000
Cache Size 20MB
Virtuozzo VM
We took a copy of our blog with its Wordpress plugins minus wp-super-cache. At any
point in time during the experiment, only one VM is up and running. The other VM is stopped.
Also, prior to running the benchmark tool, we restart the VM. Here is the base
configuration of each VM:
Dual-Core Opteron
1GB RAM
10GB Disk space
Wordpress 3.0
Unrestricted resource access (for better or worse)
Without further delay, here are the Apache Benchmark tool results for 10000 requests and 100
concurrent requests for Litespeed Web server:
View Apache Benchmark Output
Likewise, here are the Apache Benchmark tool results for 10000 requests and 100 concurrent
requests for Apache Web server configured behind Varnish:
View Apache Benchmark Output
It is important to note that the Litespeed VM ran out of memory at different times
during the experiment. The Apache node, however, did not. This means that the Apache
node was better at handling available resources. Apache was able to handle 217 requests/sec
whereas Litespeed was able to handle 111 requests/sec. Also, Litespeed took twice as long
to respond to requests. Here are some notable statistics:
Apache Benchmark
217.05 req/sec
Time taken for tests: 46.7 seconds
Transfer rate: 11177.99 [Kbytes/sec]
Failed requests: 0
Time per request: 460.7 [ms]
Litespeed Benchmark
111.15 req/sec
Time taken for tests: 89.9 seconds
Transfer rate: 183.98 [Kbytes/sec]
Failed requests: 9779
Time per request: 899.6 [ms]
The results from
Apache are impressive but needless to say that this would not
have been achievable without the help of
Varnish. Nowadays, Varnish
is a must in all high-traffic configurations. We leverage the software wherever
it makes sense for our clients.