부산대학교 도서관

Next-generation cellular networks promise to deliver immersive experiences to mobile users by seamlessly blending the virtual and physical worlds with their on-premises computing capabilities, namely connected computing capability. Critical to this promise is keeping service latency over the network consistent and low. While tremendous efforts have been made to reduce latency in networking and computing separately using various techniques such as network slicing, resource reservation, prioritization, and computation acceleration, we find that this disjointed latency management introduces non-negligible resource inefficiencies in both networking and computing, which are expected to worsen as services become more time-critical and their quality demands become heterogeneous. To address this issue, in this work, we propose a reference connected computing service model that can accommodate a variety of immersive services, and analyze how the latency components therein are intertwined. Finally, we propose a joint resource orchestration framework that can trade network resources for computing resources to minimize the total cost of operation.