Managing Call Servers for High Availability

Abstract

Achieving high availability in large scale, distributed call server systems poses challenges to automatic operation, effective load balancing and configuration. Often, the operator is required to detect outages and manually reconfigure the functioning call servers to take on the work of the failed call server or they must specify many complex contingency elements in the configuration of the collection and requires different configuration for each call server in the collection. If the failed call server was carrying a significant portion of the load of the overall system, it may not be possible to shift its entire load to a single alternate call server. Traditional solutions to this problem have been to employ an “N+1 sparing†strategy, where one additional call server is kept in reserve and is brought into service when an active call server has failed. Superclustering of call servers, supported by “territories†to encapsulate aspects of the load of a call server, addresses these problems without requiring N+1 sparing, and permits simple, yet flexible and powerful configuration of the entire system.