Drop computing: Ad-hoc dynamic collaborative computing

Abstract

Mobile applications nowadays generally consist of a frontend component running on the device, and a backend component running on the cloud that performs the larger computations. However, this usage model leads to high costs for developing the application (since a cloud infrastructure that should scale to the number of users must be maintained), and to a potentially bad user experience (if the latency is high or the users employ mobile broadband they pay for). Thus, we introduce the Drop Computing paradigm, which proposes the concept of decentralized computing over multilayered networks, combining cloud and wireless technologies over a social crowd formed between mobile and edge devices. Mobile devices and people interconnect to form ad-hoc dynamic collaborations to support the equivalent of a crowd-based edge multilayered cloud of clouds, where the capabilities of any mobile device are extended beyond the local technology barriers, to accommodate external resources available in the crowd of other devices. Thus, instead of every data or computation request going directly to the cloud, Drop Computing employs the mobile crowd formed of devices in close proximity for quicker and more efficient access. Devices in the mobile crowd are leveraged for requesting already downloaded data or performing computations, and the cloud acts as the second (or even third) option. We present a proof-of-concept implementation of Drop Computing and show, through simulations, that it is feasible for real-life usage, since it is able to drastically reduce costs while not affecting or even improving the user experience.