Abstract

Goal of the project was to design a High Availability SIP VoIP system that guarantees access to additional services like auto-attendent and automatic call distribution. The realization of the project required system to have load balancing and failover algorithms implemented. From the end users’ point of view the whole system had to figure as a single element. After taking under consideration the known methods of implementations of the HA systems, a new method has been developed and implemented. The whole system had to be immune to outage of each of the elements it consists of. There was a necessity to make the system an elastic solution with availability to add additional redundant elements in a way that required a little to no additional works with its algorithms. The last requirement was to make the system immune to the changes in its structure made by sudden outages and recovers of impacted elements. Each of the elements that recovered after the outage ought to regain all the data shared by the rest of the elements. System worked with the use of programs: Asterisk, Heartbeat and Kamailio. The whole system consists of five virtual computer platforms, set on a single physical PC. Two Kamailio servers and three Asterisk servers where the basis for the system. Kamailio was set to work as a dispatcher which assignment was to load balance the traffic between media servers. The second task of Kamailio servers was authentication of the end users. The last function was to find out outages of media servers in the system and reassign the destination media server for SIP traffic in case of its outage. Asterisk servers were the source of additional services guaranteed to the end users. Asterisk was also responsible for the dialplan for the users. Heartbeat worked on Kamailio platforms and its task was to failover between them in case of active Kamailio server’s disconnection. It also assigned virtual IP address to the active dispatcher which was used as a destination address for end users registration process. The scope of the tests undertaken for this solution was to examine the required functionality. Failover process for both Kamailio and Asterisk servers were tested, as well as load balancing algorithm. Additional tests included checking if a recently disconnected media server recovers all the data shared amongst other media servers and the communication process between system’s elements in some of the most basic SIP transactions scenarios.

Authors (3)