Exploring Computing at the Edge

Abstract

Today, mobile applications advancements have overcome limited device capabilities by offloading to costly public cloud. As the edge computing paradigm began to take precedence, a mobile device cloud (MDC) formed at the edge based on idle intra-device resources emerged. This is a result of a customized user-centric composition service request for a time-bound application. Herein, devices volunteer their intra-device resources for producing a compute environment in turn satisfying the needs of the consumer. Now, with the growth of device technology and the available interfaces for accessing multiple radio technologies, a new transport layer protocol called Multipath TCP was introduced in literature. This protocol enables multiple sub-flows to join for transmitting data simultaneously. However, in scenarios like formation of device clouds, there are issues pertaining to sub-flows that are involved in a device cloud composition. One such issue is the management of sub-flow buffer. As each of these sub-flows have their own respective buffering and characteristic delays, it leads to sub-optimal performance in term of buffer occupancy. Thereby, degrading the quality of the device cloud composition. To this end, we propose an OS side architecture that plays a crucial role in managing the traffic coming from different flows. We model an agent that works conservatively satisfying Kleinrock's law and show a proof of concept experiment.