BatTorrent: A Battery-Aware BitTorrent for Mobile Devices

Zach King Adriana Iamnitchi University of South Florida University of South Florida attempts to minimize the strain on low battery clients Computer Science & Eng. Computer Science & Eng. while retaining the benefits of “download in parallel” 4202 E. Fowler Ave. 4202 E. Fowler Ave. BitTorrent provides. Tampa, FL 33620 Tampa, FL 33620 [email protected] [email protected] Keywords Mobile content distribution, energy efficiency, Jeremy Blackburn BitTorrent University of South Florida Computer Science & Eng. Introduction 4202 E. Fowler Ave. BitTorrent depends on a tit-for-tat mechanism of Tampa, FL 33620 resource sharing to ensure fast and dependable content [email protected] distribution. In short, peers involved in a BitTorrent transfer receive data at a rate proportional to the Abstract amount of data uploaded, treating bandwidth as the BitTorrent has been shown to be an effective and only shared resource. energy-efficient means to distribute digital content. Typically, however, BitTorrent swarms are composed of While this works well for immobile peers, a mobile peer immobile peers (i.e., desktop computers) with no is also concerned with battery usage. As transmission concerns for loss of power. In the mobile realm this of data necessarily consumes more energy than concern becomes a reality. Due to the upload-in-order- reception of data, a mobile BitTorrent peer is faced with to-download nature of the tit-for-tat mechanism, a conundrum: maximize upload bandwidth to receive mobile devices must consume battery life to remain the reciprocal download bandwidth and drain battery or viable members of the swarm. We propose a novel limit the rate of data upload and suffer the solution which takes into account the variability of consequences of the tit-for-tat enforcement. Further, available battery life in a mobile BitTorrent swarm and any action the peer takes, whether it attempts to Copyright is held by the author/owner(s). minimize download time or battery consumption, can UbiComp 2009, Sep 30 – Oct 3, 2009, Orlando, FL, USA have far reaching effects on the swarm as a whole. ACM XXX-X-XXXXX-XXX-X/XX/XX. 2

BatTorrent connections will receive greater benefit than those with BatTorrent is our proposed system for dealing with weak social connections. BitTorrent swarms composed of battery-constrained peers. Our primary contribution is an adjustment to the As BatTorrent deliberately reduces the bandwidth exclusively bandwidth-based rate measure used by contribution of low capacity peers, the question of BitTorrent. The main insight of our solution relies on fairness arises. For the case of battery-constrained the observation that mobile clients share two devices, we determine that fairness is a function of the resources: bandwidth and battery. Each bit uploaded bandwidth contribution with respect to the risk of consumes not only bandwidth but also battery life. The functionality loss of a peer. I.e., a peer with low battery consumption of battery life is arguably more important is contributing more to the swarm for every bit it than that of bandwidth. Saturating upload capacity uploads than a corresponding peer with high battery. certainly prevents many operations from succeeding, but draining battery life can result in a mobile device Finally, while we expect the ubiquity of mobile, battery- becoming an expensive, shiny paperweight. constrained devices to increase dramatically in the future, it is unreasonable to expect non-constrained BatTorrent introduces a new message, BATT. The devices to vacate the scene completely. BatTorrent payload of this new message is the remaining battery makes use of non-constrained devices to offload energy capacity of the peer sending the message (over the consumption where possible. While this is the obvious interval (0, 1]). This payload is used by receiving peers solution to the trivial case where non-constrained during the choking algorithm. BatTorrent peers scale devices are able to fully saturate the download capacity the rate measure of other peers via the function f(r, of constrained devices, we have observed Wi-Fi speeds batt) = r * 1/batt, where r is the standard BitTorrent on mobile devices that would indicate a shrinking gap rate measure. The result is that peers with less battery between non-constrained and constrained device life remaining are able to scale down their upload rate speeds thus strengthening the case for BatTorrent. down in proportion to remaining battery life and not suffer choking that would normally occur. BatTorrent is implemented on the Android mobile platform and will be demonstrated during the poster As there is no reliable means to externally determine presentation. The social trust network that BatTorrent the remaining battery capacity of a connected peer, uses (GeoS) has also been submitted as a poster for BatTorrent relies on a social trust network to determine UbiComp 2009. Our poster will include preliminary the veracity of BATT messages. In short, we assume results comparing download time and battery that friends are unlikely to game each other and thus consumption of standard BitTorrent vs. BatTorrent BATT messages received from them are likely to be showing that BatTorrent enabled swarms allow low legitimate. For complex swarms involving more than battery peers to download in a reasonable time while binary social relationships the function f can be scaled minimizing impact on battery life. further via a trust level. Peers with strong social