Methods for Mitigating IP Network Packet Loss in Realtime Audio Streaming Applications

Abstract

Increasingly Internet protocol (IP)-based networks are being used for transport of realtime broadcast quality audio. IP networks provide clear benefits of cost and flexibility over circuit-switched networks. However, IP networks can have impairments that must be mitigated in order to realize the quality of service that broadcasters are accustomed to with circuit-switched networks. Among these impairments, packet loss poses the toughest challenge. In many non-realtime applications, transport protocols such as transmission control protocol (TCP) or reliable user datagram protocol (UDP) can be used to retransmit lost packets. However, for realtime media streaming applications that are delay-sensitive and where multicasting is used, such protocols cannot be used. Therefore, techniques used for packet loss mitigation have to minimize delay and work in unidirectional network paths. The challenge faced as we encounter different packet loss patterns in real networks is that the effectiveness of the mitigation techniques depends upon the model these packet losses follow. This paper first shows how packet losses in real networks can be modeled. The results of the effectiveness of different mitigation techniques are then shown, using these models and how they can be cascaded to provide for a scalable method to mitigate varying levels of packet loss.