Video streaming, in conjunction with social networks, have given birth to a new traffic pattern over the Internet: transient, localized traffic surges, known as flash crowds. Traditional traffic-engineering methods can hardly cope with these surges, as they are unpredictable by nature. Consequently, networks either have to be over-provisioned, which is expensive and wastes resources, or risk to periodically incur congestion, which infuriates customers. This demonstration shows how Fibbing can improve network performance and preserve users’ quality of experience when accessing video streams, by implementing a fine-grained load-balancing service. This service leverages two unique features of Fibbing: programming per destination load-balancing and implementing uneven splitting ratios.
Tilmans, O., Vissicchio, S., Vanbever, L., & Rexford, J. (2016). Fibbing in action: On-demand load-balancing for better video delivery. Proceedings of the 2016 conference on ACM SIGCOMM, 619-620. https://doi.org/10.1145/2934872.2959084