Hello all,
I'm guessing if OpenSER supports more than one CPU. I ask for this to know if I can achieve to double the call per second (cps) rate or registration rate using two processors instead of one.
I suppose that many of us have read the excellent "getting started 5" guide and use the scripts provided in this guide to start their first SER/OpenSER proxy. I would like to know if other users using these scripts would like to share their experience on the amount of users their OpenSER platform is able to handle, and on which hardware. It could be interesting if some users can post their OpenSER performance (cps, registration rate, number of user) with a particular "getting started" script, for example : "Authenticating ser.cfg", "call forwarding ser.cfg". Moreover, it could be great if we could think of a standard procedure to stress test the proxy. I have heard of Sipp, so do you think this program could be the base of this procedure?
I understand that OpenSER performance greatly depends of the ser.cfg, nat configuration, avpops, the database backend, but I think that it could be great to delimit some standards configurations and show what we can expect on different hardware specifications.
I'm wondering about this, cause I'd like to know if one or two OpenSER will easily support 30 000 to 40 000 users, call forwarding, redirection to voicemail; or will I have to buy a commercial product to do this.
I have a small configuration (pstn gateway ser.cfg) running with 50 users, registering every 60s and a peak of 20cps (limited by my pstn connectivity), running like a charm on a small Intel P3 700 Mhz with 256MB (it's a small test box running on my desk). If somebody could help to define a test, it will be a pleasure to share my experience on an IBM HS20 blade with a bi Xeon 3,06 Ghz and 4GB of RAM.
Thanks to all for the daily support,
rod.
On Thursday 15 June 2006 02:11, rod wrote:
Rod,
I have very little experience with extremely high loads like you're discussing. However, I do know that you could build a MySQL cluster and use a cacheless usrloc setup on each of your OpenSER servers. This would let you scale very easily; just add more servers.
Additionally, since this is cacheless, I don't think the servers will have to keep all of the users in memory, either. Could someone more knowledgeable please verify this?
---Mike
Hello all,
I'm guessing if OpenSER supports more than one CPU. I ask for this to know if I can achieve to double the call per second (cps) rate or registration rate using two processors instead of one.
I suppose that many of us have read the excellent "getting started 5" guide and use the scripts provided in this guide to start their first SER/OpenSER proxy. I would like to know if other users using these scripts would like to share their experience on the amount of users their OpenSER platform is able to handle, and on which hardware. It could be interesting if some users can post their OpenSER performance (cps, registration rate, number of user) with a particular "getting started" script, for example : "Authenticating ser.cfg", "call forwarding ser.cfg". Moreover, it could be great if we could think of a standard procedure to stress test the proxy. I have heard of Sipp, so do you think this program could be the base of this procedure?
I understand that OpenSER performance greatly depends of the ser.cfg, nat configuration, avpops, the database backend, but I think that it could be great to delimit some standards configurations and show what we can expect on different hardware specifications.
I'm wondering about this, cause I'd like to know if one or two OpenSER will easily support 30 000 to 40 000 users, call forwarding, redirection to voicemail; or will I have to buy a commercial product to do this.
I have a small configuration (pstn gateway ser.cfg) running with 50 users, registering every 60s and a peak of 20cps (limited by my pstn connectivity), running like a charm on a small Intel P3 700 Mhz with 256MB (it's a small test box running on my desk). If somebody could help to define a test, it will be a pleasure to share my experience on an IBM HS20 blade with a bi Xeon 3,06 Ghz and 4GB of RAM.
Thanks to all for the daily support,
rod.
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Hello,
On 06/16/06 15:32, Mike Williams wrote:
On Thursday 15 June 2006 02:11, rod wrote:
Rod,
I have very little experience with extremely high loads like you're discussing. However, I do know that you could build a MySQL cluster and use a cacheless usrloc setup on each of your OpenSER servers. This would let you scale very easily; just add more servers.
Additionally, since this is cacheless, I don't think the servers will have to keep all of the users in memory, either. Could someone more knowledgeable please verify this?
the CVS head has cacheless mode (3) for user location (usrloc). There are other modules that can do cache/cacheless modes - lcr, pdt, domain ... Depending on the number of the records, you can select different modes for specific modules.
Cheers, Daniel
---Mike
Hello all,
I'm guessing if OpenSER supports more than one CPU. I ask for this to know if I can achieve to double the call per second (cps) rate or registration rate using two processors instead of one.
I suppose that many of us have read the excellent "getting started 5" guide and use the scripts provided in this guide to start their first SER/OpenSER proxy. I would like to know if other users using these scripts would like to share their experience on the amount of users their OpenSER platform is able to handle, and on which hardware. It could be interesting if some users can post their OpenSER performance (cps, registration rate, number of user) with a particular "getting started" script, for example : "Authenticating ser.cfg", "call forwarding ser.cfg". Moreover, it could be great if we could think of a standard procedure to stress test the proxy. I have heard of Sipp, so do you think this program could be the base of this procedure?
I understand that OpenSER performance greatly depends of the ser.cfg, nat configuration, avpops, the database backend, but I think that it could be great to delimit some standards configurations and show what we can expect on different hardware specifications.
I'm wondering about this, cause I'd like to know if one or two OpenSER will easily support 30 000 to 40 000 users, call forwarding, redirection to voicemail; or will I have to buy a commercial product to do this.
I have a small configuration (pstn gateway ser.cfg) running with 50 users, registering every 60s and a peak of 20cps (limited by my pstn connectivity), running like a charm on a small Intel P3 700 Mhz with 256MB (it's a small test box running on my desk). If somebody could help to define a test, it will be a pleasure to share my experience on an IBM HS20 blade with a bi Xeon 3,06 Ghz and 4GB of RAM.
Thanks to all for the daily support,
rod.
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
Hello,
On 06/15/06 21:11, rod wrote:
Hello all,
I'm guessing if OpenSER supports more than one CPU.
yes
I ask for this to know if I can achieve to double the call per second (cps) rate or registration rate using two processors instead of one.
you will not double the performances for sure, but will be significant increase. There are operations that are not CPU intensive, but I/O, like retrieving/storing records in database, dns queries may slow the processing as well.
I suppose that many of us have read the excellent "getting started 5" guide and use the scripts provided in this guide to start their first SER/OpenSER proxy. I would like to know if other users using these scripts would like to share their experience on the amount of users their OpenSER platform is able to handle, and on which hardware. It could be interesting if some users can post their OpenSER performance (cps, registration rate, number of user) with a particular "getting started" script, for example : "Authenticating ser.cfg", "call forwarding ser.cfg". Moreover, it could be great if we could think of a standard procedure to stress test the proxy. I have heard of Sipp, so do you think this program could be the base of this procedure?
yes, sipp is a good tool, you can create different scenarios to test the routing. It supports many transport protocols.
I understand that OpenSER performance greatly depends of the ser.cfg, nat configuration, avpops, the database backend, but I think that it could be great to delimit some standards configurations and show what we can expect on different hardware specifications.
Some limitations come from the amount of users behind nat. If you have to proxy the RTP streams through your server, then the bandwidth may be a bottleneck.
I'm wondering about this, cause I'd like to know if one or two OpenSER will easily support 30 000 to 40 000 users, call forwarding, redirection to voicemail; or will I have to buy a commercial product to do this.
One instance of openser should be able to serve more that these numbers of users for the operations you require. The memory has to be more as you add lot of users - 4GB is more than needed.
There were performance numbers posted on the mailing list, maybe searching in the archive will reveal some.
Cheers, Daniel
I have a small configuration (pstn gateway ser.cfg) running with 50 users, registering every 60s and a peak of 20cps (limited by my pstn connectivity), running like a charm on a small Intel P3 700 Mhz with 256MB (it's a small test box running on my desk). If somebody could help to define a test, it will be a pleasure to share my experience on an IBM HS20 blade with a bi Xeon 3,06 Ghz and 4GB of RAM.
Thanks to all for the daily support,
rod.
Users mailing list Users@openser.org http://openser.org/cgi-bin/mailman/listinfo/users
-
Daniel-Constantin Mierla -
Mike Williams -
rod