I'm here at Astricon in sunny Glendale, AZ with the diverse Asterisk community talking about using SIP to improve the reliability and scalability of Asterisk.  

We had a full room this morning during our session sponsored by both ScanSource and AudioCodes, discussing different architecture tricks to improve reliability of Asterisk solutions and spread out the risk with the goal of avoiding a "career affecting" event.

A quick summary of our recommendations:

1. Use redundant and commercial grade servers to avoid simple power supply and fan failures from taking out the whole system.  
2. Use a distributed SIP architecture to separate the Asterisk server, media gateway and phone devices.
3. Implement a load-balancing scheme to spread the traffic over both Asterisk servers and dual media gateways.
4. Use Dundi to keep the dual Asterisk servers synchronized and allow them to cover for each other.
5. ...and many other helpful tips....

We had a number of interesting questions about how Asterisk handles routing the RTP streams around the application server, enabling mid-call fail-over in the case of an Asterisk crash.

For more details and/or a copy of the presentation, visit the AudioCodes micro-site on the ScanSource web site at: www.scansourcecommunications.com/audiocodes