Finally on July 22nd I had the opportunity to install the RPi in a real production environment in Germany. Since that day it is used as PBX of a non-profit organization, running a permanent installation of 7 extensions. For a period of 3 weeks following July 22nd several additional extensions were installed to support all kinds of organizational activity of a greater event. These 3 weeks gave me the opportunity to test the system running with greater load than usual. A few numbers:
|Number of extensions:||19|
|Number of trunks:||9|
|Average number of calls per day:||~60|
The RPi directly replaced a PC based Trixbox that was in service for almost 4 years. Voice quality was absolutely perfect, and overal service quality was in no way inferior to the previous Trixbox installation. Users did not notice their calls were now routed through the RPi instead, however none of them used the webinterface (which runs slower indeed compared to the old system). On peak days I recorded more than 100 calls per day. 3 concurrent calls were reported several times in the logs.
The whole installation is using SIP technology only. Most clients are AVM Fritzboxes, and several different Germany based SIP providers are used as trunks: 1und1.de, dus.net, easybell.de and partly sipkom.com
I consider the whole test a big success. However, some problems appeared as well: Directly after connecting the RPi to a PoE enabled HP V1910 switch it crashed very often. Disabling PoE on the port used by the RPi made the frequent crashes disappear, but still the device locked up every 3 to 4 days. It is hard to say what really caused this, but I am suspecting the unstable power grid at the place of installation as the source, because other devices also crashed sometimes. In order to find the cause for the crashes I will try different power supplies as soon as possible and will monitor the situation closely in the long term. The crashes need to be resolved definitely…