Ricardo Newbery ric at digitalmarbles.com
Tue Sep 25 10:52:59 CEST 2007

On Sep 25, 2007, at 1:07 AM, Dag-Erling Smørgrav wrote:

> Ricardo Newbery <ric at digitalmarbles.com> writes:
>> Varnish is good as a proxy in front of *slow* backends generating
>> dynamic content, but I suspect it's probably of marginal benefit as
>> a proxy to serve static content and possibly even when serving
>> *fast* dynamic content.
> You're wrong.  Varnish improves performance even for static content
> and fast dynamic content.  Varnish serves requests from memory, and
> doesn't spend valuable time writing logs.

If you don't need logs, it just as easy to turn this off in Apache.

>> I'm curious about the claim about Apache performance in comparison
>> to Varnish.  Is Varnish serving from cache really that much faster
>> than Apache serving static files?
> Yes, as long as your client isn't apachebench.

I'm sure it's faster, but how much faster?  Any benchmarks against a  
performance-optimized Apache?

Performance-tweaking is like security-tweaking.  At a certain point,  
it's just a black hole you can throw more money and effort into and  
get marginal returns.  And increasing performance (as in increasing  
security) often means sacrificing some power or flexibility  
elsewhere.  For some use-cases, the marginal returns and decreased  
flexibility are worth the effort.  For many, I suspect it's probably  
not.  In any case, it's better to make these decisions based on some  
objective measurements.


More information about the varnish-misc mailing list