Resource Allocation for Componentized Multimedia Service in Ubiquitous Computing Power Environment

With the rapid development of multimedia broadcasting and communication technology, multimedia services in the form of componentized network applications deployed in the ubiquitous computing environments comprised of cloud-edge-end 3 tiers of nodes are coming true. However, the existing works usually assume the application is non-partitioned or the computing environment with multiple resources is strictly hierarchical. In this work, we characterize interconnected function components in a multimedia service with a weighted directed acyclic graph (DAG), which then be accommodated with operation resources, like computing and link bandwidth, from the nodes in the ubiquitous computing environment comprised of center/edge cloud nodes and mobile devices. To solve the resource scheduling problem, which is NP-hard, we propose a reinforcement learning-based method to minimize the execution time of the service. The simulation results show that the proposed approach outperforms other baselines on saving execution time and reduce resource occupation at the same time.