Hi
I would like to have some rough estimate of the registrar performance (with DB).
If there is no figure available, does it make sense to use call performance benchmark like this one http://www.transnexus.com/White%20Papers/OpenSER_RTPproxy_test.htm and extrapolate to registrar performances ? eg. can I say that the SIP REGISTRAR is able to handle >600 registrations per second in the case of TransNexus testbed ?
Regards, Pascal
Pascal Maugeri wrote:
Hi
I would like to have some rough estimate of the registrar performance (with DB).
registrar performance heavily depends on DB mode. Write through is probably slower than write back mode.
If there is no figure available, does it make sense to use call performance benchmark like this one http://www.transnexus.com/White%20Papers/OpenSER_RTPproxy_test.htm and extrapolate to registrar performances ?
I do not think so.
regards klaus
eg. can I say that the SIP REGISTRAR is able to handle >600 registrations per second in the case of TransNexus testbed ?
Regards, Pascal
Users mailing list Users@lists.kamailio.org http://lists.kamailio.org/cgi-bin/mailman/listinfo/users
On Tuesday 21 October 2008, Klaus Darilion wrote:
I would like to have some rough estimate of the registrar performance (with DB).
registrar performance heavily depends on DB mode. Write through is probably slower than write back mode.
Some more factors this depends on:
- type of the database (mysql, postgres..) - exact setup and machine configuration (DBs likes RAM) - usage of a local DB or a remote one - exact configuration of the UA and registrar, e.g. expires values, number of DB access per REGISTER
In our experience registrar performance is not an issue. You'll probably need several machines anyway because of HA/ maintenance requirements. But because of this different factors its not easy to estimate for your special situation.
You can find some registrar benchmarks at our website: http://www.kamailio.org/docs/openser-performance-tests/
I'd suggest that you do some benchmarks with your setup, e.g. with sipp and a simple REGISTER scenario.
Cheers,
Henning
Thanks for your input ! The link you provided is very useful for that purpose. -pascal
On Wed, Oct 22, 2008 at 10:08 AM, Henning Westerholt < henning.westerholt@1und1.de> wrote:
On Tuesday 21 October 2008, Klaus Darilion wrote:
I would like to have some rough estimate of the registrar performance (with DB).
registrar performance heavily depends on DB mode. Write through is probably slower than write back mode.
Some more factors this depends on:
- type of the database (mysql, postgres..)
- exact setup and machine configuration (DBs likes RAM)
- usage of a local DB or a remote one
- exact configuration of the UA and registrar, e.g. expires values, number
of DB access per REGISTER
In our experience registrar performance is not an issue. You'll probably need several machines anyway because of HA/ maintenance requirements. But because of this different factors its not easy to estimate for your special situation.
You can find some registrar benchmarks at our website: http://www.kamailio.org/docs/openser-performance-tests/
I'd suggest that you do some benchmarks with your setup, e.g. with sipp and a simple REGISTER scenario.
Cheers,
Henning
-
Henning Westerholt -
Klaus Darilion -
Pascal Maugeri