Study Of Multi-channel Radar Parallel Signal Processing Cluster Monitoring Technology

With the development of modern information electronic war, there are the increasingdemanded functionality and performance indicators of continuous wave radar, the needs ofthe data rate and amount of computation has been greatly improved. Relying solely on theprocessing speed of CPU can not meet the needs of continuous wave radar signalprocessing, and the GPU which has powerful parallel computing capability and goodprogrammability can meet the need. This paper studies the multi-channel radar parallelsignal processing cluster technology, and sets up a server cluster based on CUDA platformfor continuous wave radar multi-channel parallel signal processing. The paper estimates thereliability and maintainability of the cluster to verify the feasibility of the cluster system.Itdesigns and implements a cluster monitoring software which is based on Ganglia. Also, itdesigns a fault disposal program of the cluster to improve the high availability of thecluster.Firstly, a multi-channel parallel signal processing cluster which is based on singlechannel continuous wave radar signal processor is designed and implemented for thereal-time measurement of target distance, speed, angle and other information. The paperestimates the reliability and maintainability of the cluster to verify the feasibility of thecluster system.Secondly, the paper introduces the Ganglia monitoring software in detail，including theconfiguration of Ganglia and the expansion of the Python indicators modules.It designs thecluster monitoring scheme of the radar signal processing cluster, including the scheme ofthe cluster monitoring software which is based on Ganglia and the scheme of the faultdisposal.Finally, the paper describes the realization process of the cluster monitoring software.Itdescribes how to deploy Ganglia in this cluster, and how to modify the configuration file tocollect the basic indicators according to project requirements. Then it describes thedeployment and writen of Python indicators modules, including GPU temperature, fiberstatus and thread status of the signal processing software. It also describes the clustermetrics data obtained, the extraction and storage. Finally, it develops a cluster monitoringinterface for signal processing results displayed and cluster status monitoring by QT.