Smooth Streaming Support for Time-Critical Streaming Media Applications

Streaming media is becoming increasingly important to many applications. Timely delivery of streaming media at the required rates is crucial to the correct functioning of applications relying on continuous input of streaming media. Compared to regular data transfer, streaming media is more sensitive to transmission delays and packet losses. Variations in transmission delays and packet losses are typically caused by network congestion which is, in turn, caused by high fluctuation in sending rates of data flows. Fluctuations in sending rate of individual data flow collectively lead to unstable network conditions, ineffective use of network bandwidth, and poor transfer quality (large delay, high data loss rate). In order to mitigate and eliminate network congestion, sending rates of data flows should be maintained as stable as possible. Fluctuation of sending rate is largely caused by transport-layer protocols. Congestion control schemes are necessary to make streaming media traffic adapt to the available network bandwidth by adjusting the sending rates of data streams. In time-critical data streaming applications, timely delivery of data is more important than reliable delivery of data. In this paper, we explore a weighted-fairness rate control method which does not rely on resource reservation or explicit congestion notifications in the network. This rate control method is the key technique in establishing an adaptive framework of timely delivery of data streams for supporting time-critical streaming media applications. This rate control scheme follows a TCP-friendly rate control (TFRC) approach. The sending rate of a data flow is adjusted based on the metrics of network congestion indication, such as the packet loss rate and round-trip time (RTT). Upon a congestion signal, the sending rate of a flow is reduced reversely proportional to the priority of the flow. Namely, the higher the priority of a flow is, the less the sending rate of the flow is reduced upon a congestion. Accurate estimation of network congestion condition related metrics is crucial to the effectiveness of this rate control method. The responsiveness of this rate control scheme to congestion largely affects the network stability. An over-responsive congestion control scheme (within one round-trip time or even less) may cause wide rate fluctuations and impede the network stability, while a slow-responsive scheme can not make data flows adapt to condition quickly. This rate control scheme also operates on metrics of congestion indication of appropriate time scale, rather than on transient congestion events.