Consume Local: Towards Carbon Free Content Delivery

Abstract

P2P sharing amongst consumers has been proposed as a way to decrease load on Content Delivery Networks. This paper develops an analytical model that shows an additional benefit of sharing content locally: Selecting close by peers to share content from leads to shorter paths compared to traditional CDNs, decreasing the overall carbon footprint of the system. Using data from a month-long trace over 3 million monthly users in London accessing TV shows online, we show that local sharing can result in a decrease of 24–48% in the system-wide carbon footprint of online video streaming, despite various obstacle factors that can restrict swarm sizes. We confirm the robustness of the savings by using realistic energy parameters drawn from two widely used settings. We also show that if the energy savings of the CDN servers are transferred as carbon credits to the end users, over 70% of users can become carbon positive, i.e., are able to support their content consumption without incurring any carbon footprint, and are able to offset their other carbon consumption. We suggest carbon credit transfers from CDNs to end users as a novel way to incentivise participation in peer- assisted content delivery.

Publication
Proceedings of the 38th IEEE International Conference on Distributed Computing Systems (ICDCS)
Aravindh Raman
Aravindh Raman
PDRA, working on network measurements
Dmytro Karamshuk
Dmytro Karamshuk
Former Postdoc, now Research Scientist at Facebook Core Data Science.
Next
Previous