GPU Accelerated Industrial Data Analysis in Private Cloud Environment

The performance of industrial and spatial data analysis is vital for decision-making systems in manufacturing sector. This paper proposes an approach to organizing a private cloud environment for the purpose of real-time analysis of data collected from industrial and measurement equipment (CNC machines, laser complexes, IoT devices and other cyber-physical system components). The approach uses sharing graphics processing unit (GPU) devices across the virtual machine. Computing working environment based on the approach is formed using the Nvidia Pascal GPU architecture in a private cloud supported by input-output virtualization technology (IOMMU). Comparative research between created computing cloud and bare metal environments is conducted and reveals the efficiency of the proposed approach. The combination of fast GPU computing and cloud environment flexibility allows simplifying the maintenance of the real time industrial analytical system. Developed solution can be used for the real time industrial system or for training their AI-based models.