|Asterisk [message #12060]
|Mon, 16 April 2007 16:01
Registered: February 2007
Asterisk dimensioning depends on your needs and also on your hardware. If you are administering your own VE's, as opposed to general hosting service with ssh access and the like, OpenVZ is well-suited to host Asterisk because of it's low impact on overall system performance (although I'm currently watching an interesting discussion going on at http://forum.openvz.org/index.php?t=msg&goto=11964&& amp;srch=wht#msg_11964 ).
BlueWave's solution involves distributing load against an open-ended number of commodity servers, but I can state that in-house testing occurs on an array of dual processor Xeon 3.0 Ghz duo-core servers each with 8GB ram. We've tested 100 VEs on a single server, each running a copy of Asterisk without problem. We haven't ventured above this mark because of ram constraints - we spec a minimum of 70MB to each instance - but with additional memory anything is possible. Concurrent calls is a separate matter altogether, but in our testing we haven't observed an impact by which VE's are participating (2 simultaneous G.711 calls in each of 100 VE's or 200 G.711 calls in a single VE). Again, we move load around based on the presence of additional server nodes in a cluster. But if your application calls for a single server, you may find this wiki entry the most helpful, it provides call-load observations on numerous types of hardware:
--- original message ---
Any users following here that have asterisk running under openvz?
Right now we have put a test box together to play with this, and we are struggling with dimensioning right now. Can anyone shed some light on parameters that worked for them, and any surrounding details, like how many extensions it handled etc.
If some of the bluewave guys could pipe in here that would be great as well.