Orchestrating heterogeneous MEC-based applications for connected vehicles

摘要

In the near future, 5G-connected vehicles will be able to exchange messages with each other, with the roadside infrastructure, with back-end servers, and with the Internet. They will do so with reduced latency, increased relia-bility, and large throughput under high mobility and user density. Different services with different requirements, such as Advanced Driving Assistance (ADA) and High Definition (HD) Video Streaming, will share the same physical resources, such as the wireless channel. Thus, a rigid orchestration among them becomes necessary to prioritize network resource allocation. This study proposes a Connected Vehicle Service Orchestrator (CVSO) which optimizes the Quality of Experience (QoE) of an in-vehicle infotainment video delivery service, while taking into account the required bandwidth for coexisting high priority services, such as ADA. To this end, we pro-vide an Integer Linear Programming (ILP) formulation for the problem of optimally assigning a video streaming bitrate/quality per user to maximize the overall QoE, considering information from the video service and the Ra-dio Access Network (RAN) levels. Our system takes advantage of recent developments in the area of Multi-access Edge Computing (MEC). In particular, we have implemented the CVSO and other service-level components and have deployed them on top of a standards-compliant MEC platform that we have developed. We exploit MEC-native services such as the Radio Network Information Service (RNIS) to offer the CVSO the necessary level of RAN awareness. Experiments on a full LTE network testbed featuring our MEC platform demonstrate the perfor-mance improvements our system brings in terms of video QoE. Furthermore, we propose and evaluate different algorithms to solve the ILP, which exhibit different trade-offs between solution quality and execution time.