Map–Reduce–Style Job Offloading Using Historical Manufacturing Behavior for Edge Devices in Smart Factory

Abstract

For smart factories in the Industry 4.0 era, edge devices can be used to run intelligent software packages (i.e., manufacturing services) to support manufacturing activities of production equipment. However, this kind of edge device may fail to provide designed functionalities in time when it encounters a sudden high-computation (SHC) job that cannot be executed by the edge device itself within a required time constraint. Thus, how to assure that an edge device can effectively execute SHC jobs so as to provide manufacturing services to equipment in time is an important and challenging issue for smart manufacturing. Exiting job-shop scheduling (JSS) methods may optimally allocate jobs for production lines but cannot directly be applied to handle the edge devices SHC jobs, which occur irregularly and are hard to specify processing time slots for JSS methods. Aimed at resolving the above-mentioned issue, this letter proposes a novel manufacturing-behavior-based map-reduce-style job offloading scheme. First, a distributed job processing architecture is designed to allow distributed edge devices to collaboratively complete an SHC job to support manufacturing. Next, a map-reduce-style program structure is developed so that an SHC job can be easily divided into multiple parts that can be executed in parallel by the selected edge devices, each device processing a part. Then, a mechanism of selecting proper edge devices to complete the SHC job, considering historical manufacturing behaviors of each edge device, is proposed for achieving high offloading efficiency. Finally, we simulate a factory with 20 machines to conduct integrated tests. Testing results demonstrate the effectiveness and excellent performance of the proposed scheme.