Retransmission schemes for streaming internet multimedia

Abstract

This paper presents a trace-driven simulation study of two classes of retransmission timeout (RTO) estimators in the context of real-time streaming over the Internet. We explore the viability of employing retransmission timeouts in NACK-based (i.e., rate-based) streaming applications to support multiple retransmission attempts per lost packet. The first part of our simulation is based on trace data collected during a number of real-time streaming tests between dialup clients in all 50 states in the U.S. (including 653 major U.S. cities) and a backbone video server. The second part of the study is based on streaming tests over DSL and ISDN access links. First, we define a generic performance measure for assessing the accuracy of hypothetical RTO estimators based on the samples of the round-trip delay (RTT) recorded in the trace data. Second, using this performance measure, we evaluate the class of TCP-like estimators and find the optimal estimator given our performance measure. Third, we introduce a new class of estimators based on delay jitter and show that they significantly outperform TCP-like estimators in NACK-based applications with low-frequency RTT sampling. Finally, we show that high-frequency sampling of the RTT completely changes the situation and makes the class of TCP-like estimators as accurate as the class of delay-jitter estimators.