Large keepalive_requests values are severely slowing-down Nginx

Question

When running a bacon (43-byte transparent pixel) load test on Nginx, we have tried several keepalive_requests values (from 10 to 100,000) and the optimal value seems to be 10.

Here are the server HTTP headers of this tiny reply:

   HTTP/1.1 200 OK
   Server: nginx/1.5.6
   Date: Wed, 23 Oct 2013 12:39:45 GMT
   Content-Type: image/gif
   Content-Length: 43
   Last-Modified: Mon, 28 Sep 1970 06:00:00 GMT
   Connection: keep-alive

Nginx is twice slower with keepalive_requests 100000 than with keepalive_requests 10.

Can you help understanding that result? Or tell what we do wrong?

For reference, here is the nginx.conf file.

Solution

Following the comment's suggestion above, I searched the Nginx Forum and the same question titled "Enabling keepalive increases request time" and posted one month ago was left without any answer.

It looks like a bug. Will have to wait for a fix.

Comments

[that guy from over there]

you'd better ask this on the nginx-forum/ml.