Web Hosting Talk : Web Hosting Main Forums : Web Hosting : Apache vs Varnish

[HostXNow]

http://loadimpact.com/result/varnish...c8e8c87a04bc81

http://loadimpact.com/result/apache....e867de28b2989e

Apache on its own really is bad.

[CodyRo]

Originally Posted by HostXNow

http://loadimpact.com/result/varnish...c8e8c87a04bc81

http://loadimpact.com/result/apache....e867de28b2989e

Apache on its own really is bad.

Varnish is a reverse proxy / caching proxy. It's not a web server / a fair comparison in the slightest.

Just to elaborate: it's caching stuff *after* it's been processed and things of that nature*. Hence the drastic differences in speed.

* Varnish is super spiffy and highly recommended, albeit a bit confusing at times if you try to get fancy (yay dynamic content in midst of static cached content).

[HostXNow]

Originally Posted by CodyRo

Varnish is a reverse proxy / caching proxy. It's not a web server / a fair comparison in the slightest.

True. Still good to compare the difference and see that Apache need to improve Apache quick.

I will be doing a proper test of LiteSpeed (OS + caching) vs Apache (OS) + Varnish (caching) soon.

[HostXNow]

As I just mentioned in another thread ... I would say only use Varnish if you have a lot of users on your site as that's where Varnish comes in, but it turns out Varnish even helps with low site activity too.

In the tests above, Varnish loads the site at 819 milliseconds for first 10 clients, where as Apache loads the site at 1.46 seconds, and the fact that Varnish page load time gets faster as more clients are loaded is just a bonus.

[Sparrow-Sean]

Varnish + LiteSpeed + Apache + eAccelerator + xCache = One big conspiracy to many.

Whatever works for an individual works, testing is ultimately going to vary under so many dominating factors - I really hate replying to these threads they are too common and all they end up in are debates! Meh!

[HostXNow]

Originally Posted by Sparrow-Sean

Varnish + LiteSpeed + Apache + eAccelerator + xCache = One big conspiracy to many.

Whatever works for an individual works, testing is ultimately going to vary under so many dominating factors - I really hate replying to these threads they are too common and all they end up in are debates! Meh!

I don't know about that. The tests look pretty good to me.

Exact same content on each dedicated server (the server used for Apache test site even has a better spec) and the LoadImpact site can be trusted. They charge 9$ day or 279$ month for paid testing services. I doubt they'd charge that kind of money and developers/testers would be upgrading if the service wasn't accurate enough.

Loadimpact looks good to me.

[Benny Kusman]

has the apache been tuned ?

[UnderHost]

Originally Posted by Benny Kusman

has the apache been tuned ?

I'm sure it wasn't, but reverse proxy/caching on static content will be always more efficiently even with apache optimized correctly.

[Serverfruit-Kris]

The Litespeed (OS + caching) and Apache (OS) + Varnish (caching) would be a great test I think! I'm looking forward for it

[Sparrow-Sean]

Mind you, these are being done via a third-party so the results may not even be accurate. It is better with internal testing and people should not hold tests so heavily on other webservers.

[Serverfruit-Kris]

Originally Posted by Sparrow-Sean

Mind you, these are being done via a third-party so the results may not even be accurate. It is better with internal testing and people should not hold tests so heavily on other webservers.

Of course network, etc. can play a role, but if you do enough testing - I think it is pretty accurate who is best. And that is why we are testing, right?

[Sparrow-Sean]

Varnish may succeed at a level, we've done testing with it and found it not to be as great as many people show it out to be, though again internal testing is probably more accurate than external services like Load-Impact.

Not saying it is bad but the results may be out of touch.

[pervez]

the views on apache & varnish can be biased, but for specific application and scenario both can have best performances..... it is requirement of application that matters.....

[Serverfruit-Kris]

Originally Posted by Sparrow-Sean

Varnish may succeed at a level, we've done testing with it and found it not to be as great as many people show it out to be, though again internal testing is probably more accurate than external services like Load-Impact.

Not saying it is bad but the results may be out of touch.

So in your internal testing, LiteSpeed performed better?