Hi Hassan, Yes, I know the details of the configuration (done the benchmarks myself). Actually we are pretty serious with these benchmarks, by default CGRateS coming with it's own stress tool (cgr-tester binary) which you can use to benchmark your own rating profiles. The machine we have used is a virtual host with a one Intel 3.4 GHz CPU and 1GB of RAM. Hardisk does not matter since we do not store anything for calculations on HDD. All the data we use in rating calculations is cached using own cache mechanism. On this machine we got the following stats (querying cost for one emulated call repeatedly -100k times) : 1. No network involved in query, (querying internally within the same process): 163655 requests per second. Applicability: CGR-SM component directly attached to FreeSWITCH with internal connection to the rater. 2. Remote RPC using GOB as transport: 12501 requests per second. Applicability: CGR-SM running on the same host with FreeSWITCH, remote rater (eg: charging for a cluster of FreeSWITCHes). 3. Remote RPC using JSON-RPC as transport (worst case scenario): 6221 requests per second. Applicability: Third party written session manager using CGR-Rater component for it's rating and implementing it's own call handling logic. I should mention that the number of requests per second should not decrease with the number of destinations you search since we use individual hashes in our cache and tested so far with 6.5 Mil. destinations without impact on performance. Happy to go deeper in your questions, not sure if appropriate to abuse here the mailing list for them, otherwise you can always join our mailing list or irc channel for a more interactive conversation. DanB PS: For the sake of correctness the screenshoot of the performance benchmarks I am talking in this email is available to be directly sent to you if interested (not accepted by the mailing list because of size). On 02.05.2014 13:43, freeswitch-users-request@lists.freeswitch.org wrote:
CGRateS seems to be a very interesting project!
The features page says "a machine of modest configuration" was used to benchmark up to 13,000 req/sec. Would you happen to know details of this "modest configuration"?
Regards HASSAN
Hi Dan, Thank you for your elaborate answer! I've subscribed to your mailing list. I'll reply using that address, so that FS list does not get unnecessarily engaged into something unrelated. Thank you once again! Regards HASSAN On Fri, May 2, 2014 at 7:27 PM, Dan Christian Bogos <danb.lists@gmail.com>wrote:
Hi Hassan,
Yes, I know the details of the configuration (done the benchmarks myself). Actually we are pretty serious with these benchmarks, by default CGRateS coming with it's own stress tool (cgr-tester binary) which you can use to benchmark your own rating profiles.
The machine we have used is a virtual host with a one Intel 3.4 GHz CPU and 1GB of RAM. Hardisk does not matter since we do not store anything for calculations on HDD. All the data we use in rating calculations is cached using own cache mechanism.
On this machine we got the following stats (querying cost for one emulated call repeatedly -100k times) :
1. No network involved in query, (querying internally within the same process): 163655 requests per second. Applicability: CGR-SM component directly attached to FreeSWITCH with internal connection to the rater.
2. Remote RPC using GOB as transport: 12501 requests per second. Applicability: CGR-SM running on the same host with FreeSWITCH, remote rater (eg: charging for a cluster of FreeSWITCHes).
3. Remote RPC using JSON-RPC as transport (worst case scenario): 6221 requests per second. Applicability: Third party written session manager using CGR-Rater component for it's rating and implementing it's own call handling logic.
I should mention that the number of requests per second should not decrease with the number of destinations you search since we use individual hashes in our cache and tested so far with 6.5 Mil. destinations without impact on performance.
Happy to go deeper in your questions, not sure if appropriate to abuse here the mailing list for them, otherwise you can always join our mailing list or irc channel for a more interactive conversation.
DanB
PS: For the sake of correctness the screenshoot of the performance benchmarks I am talking in this email is available to be directly sent to you if interested (not accepted by the mailing list because of size).
On 02.05.2014 13:43, freeswitch-users-request@lists.freeswitch.org wrote:
CGRateS seems to be a very interesting project!
The features page says "a machine of modest configuration" was used to benchmark up to 13,000 req/sec. Would you happen to know details of this "modest configuration"?
Regards HASSAN
_________________________________________________________________________ Professional FreeSWITCH Consulting Services: consulting@freeswitch.org http://www.freeswitchsolutions.com
FreeSWITCH-powered IP PBX: The CudaTel Communication Server http://www.cudatel.com
Official FreeSWITCH Sites http://www.freeswitch.org http://wiki.freeswitch.org http://www.cluecon.com
FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
participants (2)
-
Dan Christian Bogos
-
Nyamul Hassan