User Association and Behavioral Characterization during Task Offloading at the Edge

Current developments in computer vision and networking have made immersive applications, such as Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR), more affordable. As the driving force behind these types of applications is the high Quality of Service (QoS), more and more studies concentrate on offloading the application tasks to more powerful computing infrastructures without impairing the immersive user experience. This generates the problem of task offloading, defined as the transfer of resource-intensive computational tasks from a local device to an external resource-rich platform such as Cloud and/or Edge computing. Task offloading can be deemed extremely beneficial for low latency applications, however introducing several challenges in terms of task scheduling and allocation. These challenges are usually tackled via traditional optimization algorithms that can output at the same time which segments to offload and to which site (e.g. an Edge or Cloud server). These algorithms usually leverage basic input information such as task size, available computational and communication resources, etc. Going a step beyond, in this work, we propose a novel model that is able to blend the user association information through Social Network Analysis metrics and especially node centrality during the task offloading decision in an Edge infrastructure. Our results show that our approach can reduce the communication delay towards increasing the user experience.