Hi Henning,
Henning Westerholt wrote:
On Wednesday 14 March 2007 13:19, Bogdan-Andrei Iancu wrote:
Hi everybody,
based on the newly 1.2.0 OpenSER release, we ran some tests in order to give some realistic numbers about the performances of OpenSER. I think it is a very useful information for our community or other interested people. [..]
First of all congratulation to the successful release!
thanks you.
The numbers from your tests looks really good, over 8000 request per second with standard hardware is very impressive.
This is a common mistake (no offence :) ) - we got 8000 *CALLS* per second, where a call is formed of INVITE + reply + ACK + BYE + reply (see call-related section). For simple transactions for second, we got 15000 - a transaction is counted as request + reply.
I notice in several other tests this confusion about a call and transaction - and that was why we decided to do separate tests for each, to avoid any doubts about what actually was measured.
Have you run some of the tests against e.g 1.1 or 0.9.5, to see how fast is the old version in comparison to the new code?
unfortunately no, because of time restrains :(
Also, nice documentation work, it should be easy to reproduce the test with other hardware or another release.
regards, Bogdan