Design And Implementation Of A Scheduling Strategy For Parallel Processing Of Navigation Signals Based On ISSA

The research background of this paper is the project "Parallel Processing of Navigation Signals" jointly carried out by the research group and an institute.The purpose of this project is to solve the following two problems:First,in the process of processing the navigation signal data in the traditional serially implemented navigation signal processing algorithm,it takes a long time to perform the operation for the frequency points with complex operation,and the analysis of the navigation signal data may last for several days.The amount of data is large,and single-machine operation will cause the machine to be overloaded and increase the execution time of the algorithm.Second,because the load of software in the production environment fluctuates,it is difficult for front-line operation and maintenance personnel in this field to flexibly adjust the quantity of software according to the load,which affects work efficiency.However,traditional navigation signal processing algorithms can be parallelized when dealing with multiple satellites,multiple frequencies,multiple branches,and multiple evaluation items.This paper takes the parallel processing of multiple evaluation items as the main research object.Since the evaluation process of the navigation signal processing algorithm has parallelism,this paper splits the evaluation process of the traditional navigation signal processing algorithm,and combines the Docker containerization technology and the Swarm architecture to complete the algorithm parallelization.The multiple modules obtained by splitting are used for multi-machine scheduling.The main work completed in this paper is divided into two parts:navigation signal parallel processing ISSA scheduling strategy and Docker Swarm lightweight expansion and contraction module.1)The ISSA scheduling strategy for parallel processing of navigation signals,aiming at the problem of too long execution time in traditional navigation signal processing algorithms,splits each module of the evaluation process under the premise that the algorithm evaluation process is parallelizable,and combines Docker containerization technology and Swarm architecture The parallel algorithm modules are deployed on multiple machines,and the improved sparrow search algorithm is used as the scheduling strategy for multi-machine deployment,so that the algorithm modules are deployed and the computing resources of the cluster are fully utilized.When designing the scheduling strategy,firstly,the work object of the sparrow search algorithm is converted into a Docker container through population coding and population initialization.Then,through the design of the fitness function,the three factors of the Docker host’s CPU utilization,memory utilization,and the time to pull the image file are considered in the container scheduling process,and then the sparrow search algorithm population diversity and local search are considered.The problem of insufficient capacity is improved,and the ISSA scheduling strategy is finally obtained.Through the above method,the parallelization of the traditional navigation signal processing algorithm is completed,the execution time of the algorithm is reduced,and the computing resources are fully utilized.At the same time,this strategy can be used for lightweight scaling of Docker Swarm.Provide support for expansion actions in the capacity module.2)Docker Swarm Light Weight HPA module,in order to reduce the difficulty of software deployment adjustment due to load fluctuations in the production environment.In order to better scale the Swarm cluster,the docker-py function library is mainly used to complete the development of this module.This module can complete the acquisition of the resource usage of the host and Docker container in the cluster,and take the average load of the service as the level judgment criteria for expansion,then according to the expansion and shrinkage algorithm,the calculation of the change of the expansion and shrinkage capacity is completed,and finally the expansion and shrinkage action is completed to realize the adjustment of the software quantity according to the software load.The experimental results show that the ISSA-based navigation signal parallel processing algorithm scheduling strategy designed in this paper effectively improves the parallelism of the algorithm,When the sampling rate is 750Mhz,the sampling time is two seconds,and the parallelization between the evaluation items of the L2 frequency point of the GPS system reduces the running time of the algorithm from 17.803 seconds to 14.379 seconds,which is shortened by 19.3%.Compared with the default scheduling strategy in the Swarm architecture,it can effectively change the distribution of containers in the cluster and makes full use of the computing resources of the cluster;the Docker Swarm lightweight scaling module can also adjust the number of containers according to the average load of the service.