Modeling and detecting failures in next-generation distributed multimedia applications

In this paper we investigate dependability issues of next-generation distributed multimedia applications. Examples of such applications are autonomous vehicle control, tele-medicine, and audio/video control. For these applications the quality of the delivered multimedia data is a critical factor. According to the ITU-T (working group SG 12), the quality of a multimedia service as perceived by end-users is defined by three parameters: delay, delay variation, and information loss. It is paramount to formalize the concept of a failure from the user's perspective. This paper defines the correctness of a multimedia service as a function of temporal distributions of the user-related parameters. It proposes a strategy for modeling and detecting failures of the considered applications. In particular, the detection process is based on error filtering functions. We show that the combination of threshold-based mechanisms is suitable for implementing an efficient detection strategy. We also evaluate the effectiveness of the proposed mechanism both by simulations and by experiments performed on a prototype. Such a prototype is tested with respect to a case study application, consisting of distributed remote-control based on RTP/RTCP standard streaming protocols.