On 4 May 2022 16:45:31 CEST, Alan DeKok <aland@deployingradius.com> wrote:

On May 4, 2022, at 3:26 PM, Bassem Mettichi <mettichi@gmail.com> wrote:

...

i have tested freeradius V.3.0.22 and freeradius V.3.2.0 with the same configuration on centos 7 using redis as database backend and i have seen that v.3.0.22 is more performant than v.3.2.0, for a test of 25k acct/sec runned for 1hour with V.3.0.22 i didn't registered any packet lost however with V.3.2.0 i have more than 600K packet lost, i confirm that i have the same configuration and the two versions are running on the same machine with 16VCPU and 16GB RAM with localhost redis database.

There is essentially zero difference between 3.0.22 and 3.2.0 for anything related to performance. The changes are bug fixes, and updates for OpenSSL 3.

Agreed. If there really does seem to be a difference then it would be helpful to do a git bisect to find out where it happened, and also compare with v3.0.x HEAD. -- Matthew

On May 4, 2022, at 5:07 PM, Bassem Mettichi <mettichi@gmail.com> wrote

You can run tests and make comparaison, as I have made many tests and I have seen the difference between the two versions.

You aren't saying what tests you're running. Which means we can reproduce those tests. Given that (a) there's nothing in the code to cause this, and (b) we don't see anything in our tests, I think this is just random variations in the test. i.e. there may be other things happening in your network and/or OS which interfere with the tests. Alan DeKok.

Hi, Am 04.05.22 um 17:47 schrieb Bassem Mettichi:

you are not invited to reproduce the tests that i have made, you can run any performance test you want just run the same test for v.3.2.0 and other version V.3.0.X and you will see the difference.

what the heck do you expect? The FR dev team does run all kind of tests, perhaps more than you would imagine. (e.g. I like approach to compile with clang AND gcc and require consistency). Why don't you give an idea what you use for testing, e.g. eapol_test, radtest or whatever. So the community could at least see what FR modules are used in those tests. If you have a real performance issue with FR _itself_, you are most probably a real bit Telco with millions of clients -- are you? And even then, the first thing to do would be to isolate the internal FR processing from database access (LDAP, SQL...) and other external dependencies such as MS AD access or whatever. Otherwise, you are _most_ probably recording some indirect effect of some external dependency which slows down your response times. Cheers, Martin -- Dr. Martin Pauly Phone: +49-6421-28-23527 HRZ Univ. Marburg Fax: +49-6421-28-26994 Hans-Meerwein-Str. E-Mail: pauly@HRZ.Uni-Marburg.DE D-35032 Marburg

Hello, OS: centos 7 ( 16 VCPU, 16GB RAM) Freeradius: complied with GCC 4.8.5 Database: Redis Test Tool: Radperf i have tested only accounting requests ( start ) Redis and Freeradius are installed on the same server. Thanks Bassem Le sam. 7 mai 2022 à 00:27, Alan DeKok <aland@deployingradius.com> a écrit :

On May 4, 2022, at 11:47 AM, Bassem Mettichi <mettichi@gmail.com> wrote:

...

you are not invited to reproduce the tests that i have made,

That's ridiculous.

"I did stuff. But I'm not going to tell you what it is. You have to figure it out. Oh, and it's broken, so you have to fix it".

...

you can run any performance test you want just run the same test for v.3.2.0 and other version V.3.0.X and you will see the difference.

If you're not going to give any useful information, I suggest that you stop posting messages.

Alan DeKok.

- List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html