Of Pingtel's three open source IP PBXs, we tested SIPxchange Enterprise Communication Server, Version 3.7. Pingtel also has the SIPxNano for fewer than 30 users, as well as a free, downloadable version of its sipX product, which is supported by the SIPfoundry open source community.

We installed SIPxchange from a single CD that included the Linux CentOS 4 and the IP PBX software. The installation requires the administrator to enter information about 15 system and network parameters using a basic GUI; the system was up and running in 15 minutes. A second system was brought online to test failover support.

The core infrastructure of the Pingtel system is completely rooted in SIP, and can be installed on multiple Linux kernels or ported to Sun Solaris. No media streams pass through the SIPxchange server. RTP streams are passed directly between endpoints with just call-control messages relayed to the server.

This allows for reduced workload requirements on the IP PBX system, which gives Pingtel an architectural advantage over most. The other servers included with the product -- such as authentication, registration and presence servers -- are software stacks running on the same physical box but are isolated from the IP PBX from a software perspective. So if one of them has issues, it does not bring down the others.

The administrative Web interface allows for quick and easy setup and management. New users can be provisioned and configured separately or as a group. For large numbers of new users, the interface's support for a CSV format lets you import settings from an Excel spreadsheet. The proper format for the CSV file is included on the "add users" interface.

Because SIPxchange is exclusively SIP-based, adding gateways and other SIP devices is as easy as adding a user to the system. Outbound and inbound call route plans can then be developed from the default settings and used.

For high availability the SIPxchange system can be configured for load balancing and redundancy with multiple servers in multiple locations. When we simulated a catastrophic failover, we found that as long as the endpoint being used supported more then one DNS SRV record, it was able to fail over automatically. Because the calls handled by the Pingtel system are randomly load-balanced, service was not interrupted for some of the endpoints that were registered to the failover call controller.