1. Brekeke Product Name and version:
2.4.8.6
2. Java version:
1.6.0_25
3. OS type and the version:
linux centos 5.5
4. UA (phone), gateway or other hardware/software involved:
zoiper, xlite, vippie
5. Select your network pattern from http://www.brekeke-sip.com/bbs/network/ ... terns.html :
Pattern 9
6. Your problem:
I am new to Brekeke sip server. At this moment i am working on Evaluation copy of your product and a few question have emerged
1. What hw description do you suggest for 3000 concurrent calls, 80 CallPerSeconds?
Please assume that 90% of all traffic will be pure class4 sip trunks authorized by ip with an option of full RTP Proxy and without.
2. Is there any other redundancy scenarios apart from one Primary one Secondary taken into account?
If not, what about placing opensips before the call reaches two Brekeke "clusters", both containing 1+1.
Opensips could do load balance all the calls between cluster_1 and cluster_2.
Are you okay with this or have better/other wayout?
3. Is there any way to troubleshoot Brekeke apart of heaving net.registrar.loglevel.file/net.sip.loglevel.file.
We would like to debug why UA couldn't be registered, why dial plan was not chosen and so on, so to see much more that just sip traces.
It could really speed up to troubleshoot many problems.
4. We have spotted one strange behavior of Brekeke server.
Lets imagine there is one call placed and we need to restart the server asap.
Of course we are notified that all the calls will be teared down which is okay.
Server is sending BYE to UAs but no cdr is created for these calls.
We cannot see them in Call Logs.
Could you please point us where the problem might be located?
I look forward to your reply.
A new fan of Brekeke sip server
Moderator: Brekeke Support Team
> 1. What hw description do you suggest for 3000 concurrent calls, 80 CallPerSeconds?
I can recommend that you use a hardware which has a quad-core or higher CPU if you are planning to provide a service.
Note that network-bandwidth should be wide enough because you are planning to handle RTP traffics.
> 2. Is there any other redundancy scenarios apart from one Primary one Secondary taken into account?
A software HA such as Marathon will be a solution..
http://www.marathontechnologies.com/
> Opensips could do load balance all the calls between cluster_1 and cluster_2.
Why do you use Opensips for load-balancing?
Have you looked at the following wiki topic?
http://wiki.brekeke.com/wiki/Brekeke-Load-Balancing
3. Is there any way to troubleshoot Brekeke apart of heaving
The log is not just SIP trace if you set correct logging parameters.
It will help your troubleshooting.
> Server is sending BYE to UAs but no cdr is created for these calls.
CDR will be updated after BYE timeout happens because the SIP server are waiting for a response from UA in the above case.
I can recommend that you use a hardware which has a quad-core or higher CPU if you are planning to provide a service.
Note that network-bandwidth should be wide enough because you are planning to handle RTP traffics.
> 2. Is there any other redundancy scenarios apart from one Primary one Secondary taken into account?
A software HA such as Marathon will be a solution..
http://www.marathontechnologies.com/
> Opensips could do load balance all the calls between cluster_1 and cluster_2.
Why do you use Opensips for load-balancing?
Have you looked at the following wiki topic?
http://wiki.brekeke.com/wiki/Brekeke-Load-Balancing
3. Is there any way to troubleshoot Brekeke apart of heaving
The log is not just SIP trace if you set correct logging parameters.
It will help your troubleshooting.
> Server is sending BYE to UAs but no cdr is created for these calls.
CDR will be updated after BYE timeout happens because the SIP server are waiting for a response from UA in the above case.
Is Quadcore, Intel Xeon E5620 2.4GHz, 8GB RAM enough for these figures?Harold wrote:> 1. What hw description do you suggest for 3000 concurrent calls, 80 CallPerSeconds?
I can recommend that you use a hardware which has a quad-core or higher CPU if you are planning to provide a service.
Note that network-bandwidth should be wide enough because you are planning to handle RTP traffics.
Is it really enough for full RTP proxying?
Right additional Brekeke sip server might be used for this case.> Opensips could do load balance all the calls between cluster_1 and cluster_2.
Why do you use Opensips for load-balancing?
Have you looked at the following wiki topic?
http://wiki.brekeke.com/wiki/Brekeke-Load-Balancing
Do you mean setting "log4j.logger.com.brekeke.util.Dump=DEBUG, RLOG" ?3. Is there any way to troubleshoot Brekeke apart of heaving
The log is not just SIP trace if you set correct logging parameters.
It will help your troubleshooting.
I already have two entries in Configuration>Advanced:
net.registrar.loglevel.file=255
net.sip.loglevel.file=255
and it produces only SIP traces.
Not exactly. SIP server is sending BYE to both UAs, so it knows where the call is ended. There are no further BYEs, becuase the call was teared down properly by UAS.> Server is sending BYE to UAs but no cdr is created for these calls.
CDR will be updated after BYE timeout happens because the SIP server are waiting for a response from UA in the above case.
> Is Quadcore, Intel Xeon E5620 2.4GHz, 8GB RAM enough for these figures?
I think No Problem.
> Is it really enough for full RTP proxying?
It will be depending on network-bandwidth because all of RTP traffics are exchanged through the SIP server.
Try the following site.
http://www.bandcalc.com/
> Do you mean setting "log4j.logger.com.brekeke.util.Dump=DEBUG, RLOG" ?
What kind of information do you need from log??
You can customize log data with DialPlan.
> There are no further BYEs, becuase the call was teared down properly by UAS.
Even so, the SIP server will write a CDR record after timeout.
I think No Problem.
> Is it really enough for full RTP proxying?
It will be depending on network-bandwidth because all of RTP traffics are exchanged through the SIP server.
Try the following site.
http://www.bandcalc.com/
> Do you mean setting "log4j.logger.com.brekeke.util.Dump=DEBUG, RLOG" ?
What kind of information do you need from log??
You can customize log data with DialPlan.
> There are no further BYEs, becuase the call was teared down properly by UAS.
Even so, the SIP server will write a CDR record after timeout.
> Is it really enough for full RTP proxying?
It will be depending on network-bandwidth because all of RTP traffics are exchanged through the SIP server.
I was rather thinking about hardware not bandwidth consumption.
Is Quadcore, Intel Xeon E5620 2.4GHz, 8GB RAM enough for full RTP on Brekeke with 3000cc 80CPS referring to your experience you have another hw proposal?
> Do you mean setting "log4j.logger.com.brekeke.util.Dump=DEBUG, RLOG" ?
What kind of information do you need from log??
You can customize log data with DialPlan.
There will be fine if i could find sth similiar information to the onces in opensips. So a kind of logic behavior, why this dial plan was chosen, the other not, how registration process looks like while UA sends REGISTER, what happened with BYE but not in signalization level and so on.
Could you please point me to some examples how to implement logs with Dialplan or post simple example please.
> There are no further BYEs, becuase the call was teared down properly by UAS.
Even so, the SIP server will write a CDR record after timeout.
What kind of delay are you talking about?
20mins have passed and still nothing.
Thanks in advance,
Maciej
It will be depending on network-bandwidth because all of RTP traffics are exchanged through the SIP server.
I was rather thinking about hardware not bandwidth consumption.
Is Quadcore, Intel Xeon E5620 2.4GHz, 8GB RAM enough for full RTP on Brekeke with 3000cc 80CPS referring to your experience you have another hw proposal?
> Do you mean setting "log4j.logger.com.brekeke.util.Dump=DEBUG, RLOG" ?
What kind of information do you need from log??
You can customize log data with DialPlan.
There will be fine if i could find sth similiar information to the onces in opensips. So a kind of logic behavior, why this dial plan was chosen, the other not, how registration process looks like while UA sends REGISTER, what happened with BYE but not in signalization level and so on.
Could you please point me to some examples how to implement logs with Dialplan or post simple example please.
> There are no further BYEs, becuase the call was teared down properly by UAS.
Even so, the SIP server will write a CDR record after timeout.
What kind of delay are you talking about?
20mins have passed and still nothing.
Thanks in advance,
Maciej
> Quadcore, Intel Xeon E5620 2.4GHz, 8GB RAM enough for full RTP on Brekeke with 3000cc 80CPS referring to your experience you have another hw proposal?
I think you need to use more powerful hardware.
Or.. you need to use multiple Brekeke SIP Server to divide traffics.
> why this dial plan was chosen, the other not, how registration process looks like while UA sends REGISTER, what happened with BYE but not in signalization level and so on.
Read sv.xxxx.log.
For REGISTER, set "net.registrar.loglevel.file=255".
I think you need to use more powerful hardware.
Or.. you need to use multiple Brekeke SIP Server to divide traffics.
> why this dial plan was chosen, the other not, how registration process looks like while UA sends REGISTER, what happened with BYE but not in signalization level and so on.
Read sv.xxxx.log.
For REGISTER, set "net.registrar.loglevel.file=255".