Method for realizing high-speed and safe server cluster management through FPGA virtualization deployment

Abstract

The invention discloses a method for realizing high-speed and safe server cluster management through FPGA virtualization deployment, and belongs to the field of high performance data processing, whichcan improve the speed, the reliability and the safety of services such as access, programming and calculation of small and medium-sized enterprise server clusters for externally deploying FPGA resources a. The method is characterized in that a Docker container engine and a Kubernetes cluster container orchestration engine management system are adopted to construct FPGA virtualization deployment;available virtualized FPGA resources are automatically provided for users outside the cluster in a micro-service mode; then, when a user accesses the virtualized FPGA container, the user can obtain access permission to all resources of FPGA hardware, that is, an operating system kernel of the host machine is directly used to process mass data, and after the user finishes using, the system can automatically release and recycle the resources for next resource allocation and use.

Description

technical field [0001] The invention relates to the field of high-performance data processing, in particular, using FPGA virtualization deployment and scheduling methods to operate laboratories or small and medium-sized enterprise server cluster management services. Background technique [0002] Nowadays, Internet applications generate massive amounts of data, and the demand for computing in data centers is also rising rapidly. The computing demands of various applications such as deep learning online prediction, video transcoding during live broadcast, image compression and decompression, and HTTPS encryption have far exceeded the capabilities of traditional CPU processors. [0003] In the past, semiconductor technology continued to evolve rapidly, and processor performance doubled every 18 months. As semiconductor technology gradually approached the physical limit, Moore's Law was almost invalidated, circuits became more and more complex, and the development cost of each d...

AI Technical Summary

Problems solved by technology

The current application method is that users directly perform high-performance computing on the physical machine equipped with FPGA, which causes a series of problems. Users may make misoperations due to lack of technical experience, which directly causes the physical machine to fail to work normally, and even endangers system security; many Multiple users occupying the same physical machine at the same time may lead to serious resource competition and affect efficiency; users log in to the FPGA server and perform operations by assigning IP addresses, leaving traces of operations on the physical machine, thereby making the life cycle of the entire FPGA resource management (allocation, recycling and reuse) become extremely inefficient

Images