Multi-scale Validation of Structural Models of Audio Traffic

This paper identifies the structural properties of RealAudio traffic, develops an application-level model and validates the model using multi-scale analytic techniques. RealAudio traffic is quite different from web traffic and constant bit rate traffic, models commonly used for Internet and multimedia traffic respectively today. We demonstrate that RealAudio has rich behavior across a range of time scales, with regularity at large time scales, strong periodicities from RTT to a few seconds, and complex behavior at small time scales. A simulation model was developed which captures the main structural characteristics of RealAudio, and demonstrates the importance of multiscale analysis in validating this model. We also demonstrate via simulation that RealAudio loss rates can be noticeably reduced by slightly changing its characteristics to reduce burstiness. Although our results focus on Real Audio traffic, this methodology may be useful for studies of other types of real-time streaming media.