Abstract
Cloud manufacturing is a service-oriented networked manufacturing paradigm, which integrates manufacturing resources and capabilities, to enable the intelligence and digitization of the whole life cycle from design to scrap of products. The manufacturing process of complex products requires the collaboration of a large number of heterogeneous devices, the devices accessed to a manufacturing cloud can generate a bulk of data and computing requirements. Traditional cloud computing can not meet the demand for timely and efficient task processing anymore. Edge computing is a computing paradigm that enables tasks to be processed close to the edge, in order to reduce the load on the cloud, and enhances the overall responsiveness of cloud manufacturing systems. However, the computing performance and computing resources on edge nodes are limited, which cannot meet the complex computing tasks. In this paper, a modeling method of task scheduling for FPGA-based edge computing framework is proposed to load balance the edge computing network by dynamic task scheduling and algorithm hardware-based acceleration. This framework builds a task model to describe task information. The task offloading rules are decided by task information and edge nodes states, then task data is subsequently sent to the target edge node. FPGAs are introduced in the edge nodes to enhance the computing performance, which avoids the task can not be processed due to the insufficient processing capacity of a single node, in order to make the cloud manufacturing system load balance. Finally, a case is passed to verify that the proposed framework can handle the task in a timely and efficient manner.
