Real-time performance evaluation of media pipeline plug-in architectures

Abstract

Deployment of real-time applications has become ubiquitous in recent years. These applications (e.g. streaming video over the Internet) however, are processor intensive and require fast video-processing techniques for more flexible, efficient operation. Media pipeline plug-in architectures have been introduced to provide these capabilities in form of extensible, integrated frameworks. Several architectures are in existence and many future designs are expected. This paper serves to design a set of measurement metrics on which to base the comparative performance analysis of the real-time capabilities of two of the architectures; DirectShow and GStreamer. This exemplifies the decision process through which a particular architecture can be isolated as that best suited for a specified real-time application. The metrics used are: average processing speed, plug-in scalability, threading overhead and programming complexity. Based on these metrics, effective comparison of other plug-in architectures can be made to determine their real-time performance in relation to a particular application.