(A) framework for performance measurement of signaling protocols

Abstract

This paper presents a framework for performance measurement of signaling protocol servers in VoIP (Voice-over-IP). The framework spans all aspects of performance measurement, from protocol description to analysis of results. Previous testers have provided users with limited flexibility in supported protocols, test scenarios and modification of protocols. To overcome the limitation, a new measurement framework was proposed. The framework provides a general description platform so that any signaling protocols can be tested within the framework. The platform was derived from the characteristics of signaling protocols and a real measurement situation in VoIP. In addition, test scenarios can be described and modified as well, thus ensuring a wide range of test scenarios and performance indices. When describing a protocol’s behavior, two formats are used; a tabular format for description and a graph format for display. Through the formats, the framework provides an easier description and debugging environment than previous protocol specification languages. The framework consists of four parts: protocol description, test scenario description, performance measurement and analysis of results. In the protocol description step, a protocol’s behavior is described using the DEVS formalism. Then, a test scenario is set as a projected model of the full behavior description according to the performance metric to be measured. During performance measurement, a real measurement situation with multiple simultaneous users and successive calls is automatically implemented. At the analysis stage, raw measurement samples are converted into statistical data through Excel automation. The framework was applied to measure the performance of a SIP (Session Initiation Protocol) proxy server and a MGCP (Media Gateway Control Protocol) media gateway controller. The results have proven accuracy in measurement result and flexibility in measurement scenarios and performance metrics.