Principles of Scaling Cloud Computing of Digital Twins Using Accelerator Computers with Mass Multi-Core Parallelism

In relation to large industrially significant systems (using the example of nuclear power plants (NPP)), the ways of developing the architecture and functionality of cloud systems are considered. The development ways are aimed at providing full-featured cloud services based on industry standards for a system-holistic solution to the entire range of problems of using digital twin technologies with high computational complexity. The possibilities of increasing computing performance through the hybrid amplification of cloud servers by interfacing them with accelerator computers with mass, dynamically scalable multi-core parallelism (hundreds, thousands and more computing cores) are shown. The method of dynamic scaling of mass multi-core parallelism is proposed, intended to expand the space of virtualization of cloud resources.