Resource allocation for multimedia traffic flows using rate variance envelopes

In order for networks to support the delay and loss requirements of interac tive multimedia applications, resource management algorithms are needed tha t efficiently allocate network resources. In this paper, we introduce a new resource allocation scheme based on rate variance envelopes. Such envelope s capture a flow's burstiness properties and autocorrelation structure by c haracterizing the variance of its rate distribution over intervals of diffe rent length. From this traffic characterization, we develop a simple and ef ficient resource allocation algorithm for static priority schedulers by emp loying a Gaussian approximation over intervals and considering a maximal bu sy period. Our approach supports heterogeneous quality-of-service requireme nts via our consideration of prioritized service disciplines, and supports heterogeneous and bursty traffic flows via our general framework of traffic envelopes. To evaluate the scheme, we perform trace-driven simulation expe riments with long traces of compressed video and show that our approach is accurate enough to capture most of the available statistical multiplexing g ain, achieving average network utilizations of up to 90% for these traces a nd substantially outperforming alternate schemes.