Video: Procrastinator: pacing mobile apps' usage of the network
There are two emerging trends in the mobile data world. First, mobile data is exploding at a rapid rate with analysts predicting 25-50X growth by the year 2015. The second trend is that users are demanding greater degree of flexibility in selecting their operators at fine timescales. Across Asia, dual-SIM phones have become popular, while Apple is rumored to be designing a Universal SIM that will allow iPhone users to toggle between different operators. This latter trend points towards an impending disruption in wireless service models which could also be the need of the hour from the spectrum shortage perspective. This points towards a new service model where users can choose an operator based on application needs. However, if users make this choice greedily without network assistance, it can exacerbate spectrum scarcity and degrade user experience. In this work, we consider user devices with multiple network interfaces (3G, LTE etc.) that can be simultaneously active and each running multiple applications. We propose the MOTA service model to enable users to associate each interface with the operator of choice at fine time scales. Under the MOTA service model, through concise signalling information, operators provide information about their own network, so that each user can (i) choose a suitable operator for each interface, and (ii) choose an interface for each active application. We make the following contributions in this paper. First, we propose concise network signalling that assists users to make informed choices even under mobility. Second, we develop user-choice algorithms that maximize a suitable notion of user satisfaction while using spectrum resources efficiently. Third, we perform extensive evaluation over actual base station deployment in a city coupled with real signal propagation maps. Our results with two operators show that, MOTA service model provides capacity gain in the range 2.5-4X over the current existing service model. Finally, we argue that our solution is practically implementable by combining appropriate IEEE standards and IETF proposals.
