How does FastCGI work under the covers?

[Maxim Dounin]

Hello!

On Sun, Jun 30, 2013 at 05:10:23PM -0400, kgk wrote:

> Hello Everybody,
> 
> I have a simple question about the Nginx FastCGI implementation.  Let's say
> that 1,000 web browsers are requesting pages from our server, and we're
> using FastCGI within Nginx.
> 
> While the FastCGI specification permits sending little pieces of each
> request ("multiplexing" them), my understanding is that most (all?)
> implementations don't do this.  Therefore,
> 
>  (1) does Nginx FastCGI provide exactly one complete HTTP request at a
> time?
> 
>  (2) does Nginx FastCGI want us to respond to that particular HTTP request
> before it gives us the next one?
> 
> The reason that I ask is this:  If my FastCGI server is high-latency, I'm
> wondering if it can get lots of requests to work on, or if it will have to
> finish one and only then get the next one, and only Nginx knows the answer
> to this question.  :)

FastCGI multiplexing isn't used by nginx.  That is, within a 
single connection to a fastcgi application only one request is 
sent, and then nginx will wait for a response.  More connections 
will be opened if there are multiple simulteneous requests.

-- 
Maxim Dounin
http://nginx.org/en/donation.html