> I'll do it but I guess the test will no longer be so relevant because I want
> to simulate different users.

Real user/browser DO keep-alive. Sendings thousands of requests per
second in dedicated TLS session is not what you would see in real life
from real users.



> Anyway, the question is in fact: is this http/https ratio normal?
> If is it than I'll live with it, but if not then I might do something wrong.

Your test (small files, each file in a new TLS session) is basically measuring
the amount of TLS handshakes per second your CPU can manage, nothing
else.

Therefor, you don't have a "ratio" here, as that would require a realistic
test settings.

Enabling keepalive on ab is one of the things you can do. I don't know
ab, so not sure if there is a better way. I also do not know if ab supports
SSL session caching or TLS tickets, which you would have to keep in
mind when benchmarking.

The bottom point is, you need to understand your benchmark before
you try to understand its result.
_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx