Fog computing offers increased performance and efficiency for Industrial Internet of Things (IIoT) applications through distributed data processing in nearby proximity to sensors. Given resource constraints and their contentious use in IoT networks, current strategies strive to optimise which data processing tasks should be selected to run on fog devices. In this paper, we advance a more effective data processing architecture for optimisation purposes. Specifically, we consider the distinct functions of sensor data streaming, multi-stream data aggregation and event handling, required by IoT applications for identifying actionable events. We retrofit this event processing pipeline into a logical architecture, structured as a service function tree (SFT), comprising service function chains. We present a novel algorithm for mapping the SFT into a fog network topology in which nodes selected to process SFT functions (microservices) have the requisite resource capacity and network speed to meet their event processing deadlines. We used simulations to validate the algorithm's effectiveness in finding a successful SFT mapping to a physical network. Overall, our approach overcomes the bottlenecks of single service placement strategies for fog computing through composite service placements of SFTs.

翻译：雾计算通过在传感器附近进行分布式数据处理，为工业物联网应用提供了更高的性能和效率。鉴于物联网网络中的资源限制及其竞争性使用，现有策略致力于优化选择哪些数据处理任务应在雾设备上运行。本文提出了一种更有效的数据处理架构以实现优化目标。具体而言，我们考虑了物联网应用为识别可操作事件所需的三项核心功能：传感器数据流处理、多流数据聚合与事件处理。我们将该事件处理流程重构为一种逻辑架构——服务功能树，该架构由服务功能链组成。我们提出了一种新颖算法，用于将服务功能树映射到雾网络拓扑中，确保被选中处理服务功能树功能（微服务）的节点具备必要的资源容量和网络速度，以满足事件处理的截止时间要求。通过仿真实验验证了该算法在实现服务功能树到物理网络成功映射方面的有效性。总体而言，我们的方法通过服务功能树的复合服务部署，克服了雾计算中单一服务放置策略的瓶颈。