PULSE - an adaptive practical live streaming system

Pianese, Fabio

Live Streaming consists in distributing live media (video and audio) to large audiences over a computer network. Providing a live streaming service over the Internet presents many challenges: the application must respect the timing and quality constraints imposed by the nature of live media and by user expectations while struggling with the practical limitations due to the best effort properties and unpredictable dynamics of the Internet. Because of the limited deployment of native IP multicast, an Internet-based live streaming application with a global scope can only rely on end-to-end network primitives, such as unicast connections. The traditional client-server approach to live streaming has a serious scalability limit, as the upload capacity requirement at the server grows linearly with the user population. A P2P solution has the big advantage of seamlessly scaling to arbitrary population sizes, as every node that receives the video, while consuming resources, can at the same time offer its own upload bandwidth to serve other nodes. In theory, if every node contributed on average at least as much as it consumed, the P2P system would have enough resources to grow indefinitely. This thesis presents and evaluates PULSE, a practical P2P live streaming system intended for large-scale deployment over the Internet. PULSE uses an unstructured mesh-based design and relies on local pairwise incentives as its peer selection mechanism. The most innovative feature of PULSE is the unique coupling of incentives with feedback derived from data reception, which leads to the emergence of clusters that regroup nodes with similar resources. By exploiting this intrinsic clustering phenomenon and by leveraging latency measurements to estimate network locality, PULSE is capable to successfully operate in a wide range of resource-constrained real world scenarios and to support dynamic user populations and heterogeneous node upload capacities

Digital Security
Eurecom Ref:
© Université de Nice. Personal use of this material is permitted. The definitive version of this paper was published in Thesis and is available at :
See also:

PERMALINK : https://www.eurecom.fr/publication/2408