Skip to Main Content
We discuss the dynamic scaling of distributed session control servers. Dynamic scaling refers to changing the system configuration by servers dynamically joining or leaving. In session control servers, availability and real-time processing must be guaranteed. Therefore, dynamic scaling must be achieved while ensuring these requirements. We propose a cluster model that makes it possible for session control servers to achieve both scalability and dynamic scaling. One problem when a node joins a server cluster is that the lookup load that occurs when the new node takes over call states makes it impossible to continue processing. Consequently, this overload can reduce service reliability. We propose two methods for reducing the overload and describe their properties. We evaluated the methods in a set of experiments. The evaluation using a Session Initiation Protocol application revealed that dynamic scaling was efficient, which indicates the methods are effective in reducing look-up load.
Date of Conference: 5-9 Nov. 2012