Sunday, January 21, 2007

Kamailio (OpenSER) vs SER performances

Following lot of questions and comments related to the results published by Iptelorg guys related to openser vs ser performances on their site, I will give some clarifications.

First, the tests were done only by Iptelorg guys. They didn't ask anybody of openser development team if could help them to make accurate tests and make sure that openser and ser have same internal parameters. For people familiar with these SIP servers, it is well known that the applications uses internal hash tables, different number of processes, memory sizes, and so on -- this parameters can influence a lot if they have different values.

Because they continued saying tests are fair and reflect reality, we had to investigate and debug ser which ended in something pretty much expected -- they did performance improvement, but the processing was totally wrong - we reported the bug, and more or less it was admitted that it was the source of the performance improvement in ser stable version (although some from Iptelorg tried to say the results of the bug were not critical -- but was a bug). Follow the thread on mailing list for more details:

Removing that bug, ser gets back more or less to same performances.

Kamailio (OpenSER) will do performance measurements in the testing period for version 1.2.0. But it will be only for openser, as we won't try to claim we know ser that well to be sure that any eventual comparison is fair.

In open source it is expected to have friendly relationship with similar projects, not unsustainable accuses and unfair approaches and comments about the others.

No comments:

Post a Comment