Modelling the offload of AI Tasks in Mobile Clouds

In this paper, we systematically model the performance of running AI inference tasks on Mobile Cloud (MC) systems. Mobile devices collect the monitoring data and perform the model inference tasks. When the mobile devices cannot respond timely, the mobile device offloads a portion of inference tasks to the cloud server. We aim to model the performance of tasks running in such a MC system and also the resource capacities that the cloud server must have to achieve the required task performance.