Research On Dynamic Reconfgurable Resource Scheduling And Kubernetes Task Scheduling Of Remote Interactive FPGA Experiment Platform

The remote interactive FPGA experimental platform provides a convenient experimental environment for universities,which not only brings students a fexible way to get educated but also facilitates teachers from diferent universities to share teaching experiences with each other,especially during the COVID-19 period,using the remote interactive FPGA experiment platform to assist teaching can solve the pain point that ofine experiment work is diffcult to carry out and bring great convenience for teaching work.However,due to students’ exclusive use of FPGA resources and the volatility of the actual network deployment environment,the platform has problems with low FPGA resource utilization and insuffcient experimental stability.In order to guarantee the smoothness and effciency of the experiment,this paper studies the dynamic reconfgurable resource scheduling strategy and cloud task scheduling strategy of the remote interactive FPGA experiment based on the teaching experiment scenarios.Besides,this paper builds a platform operation example to test its operation performance.The main work of this paper is as follows:(1)Research on FPGA dynamic reconfgurable resource scheduling strategy based on the RDRR algorithm.Aiming at the problem that the existing Round-Robin scheduling algorithm and its variant algorithms cannot dynamically adjust the time slice,this paper proposes the design of a dynamic time slice and the dynamic weight,which allocates diferent time slices according to diferent task states.When the length of the remaining operation steps of the task is short,the remaining length optimization mechanism proposed in this paper starts to optimize it,which prevents the task from entering the next scheduling round and causes the waiting time to increase so that the system throughput is improved.(2)Research on Kubernetes scheduling strategy based on the QNS-IMPSO algorithm.Aiming at the problem that the default scheduling strategy of Kubernetes cannot meet the scheduling requirements of complex application scenarios,this paper frst sets restrictions on Qo S and node stability parameters so that the experimental resources with poor status can be fltered out in the pre-selection stage,and the stability of task execution can be guaranteed;Secondly,aiming at the problem of premature convergence of the existing particle swarm optimization algorithm and its variant algorithms,this paper proposes a multi-swarm particle communication mechanism based on random selection and a particle mutation disturbance mechanism to expand the particle search range and prevent the algorithm from falling into a local optimum.(3)Instance construction and performance analysis of remote interactive FPGA experiment platform.This article frst designs the overall platform architecture and formulates platform instances to achieve the goal.Secondly,this article builds a remote interactive FPGA experiment platform instance in a real scene,tests the platform’s operating performance,and analyzes its performance according to the number of requests,number of task executions,and other indicators.The running result shows that the platform performs well in actual scenarios.