Research And Implementation Of Pulsar Computing And Search Service Platform

The collection and processing of astronomical data is an important driving force for the development of astronomy and an organic combination of theoretical verification and unknown exploration.Tianyan,China's Five Hundred Meters Aperture Telescope(FAST),has made outstanding contributions to astronomical data observation.At present,Key Laboratory of Information and Computing Science of Guizhou Province,Guizhou Normal University,undertakes part of the calculation of FAST data.The pulsar search computing cluster in the laboratory calculates pulsars,single pulses,FAST radio bursts and other data.Based on FAST's huge data volume,the laboratory pulsar search computing cluster has been upgraded and iterated in terms of hardware and software,with a total of about 150 computing nodes participating in the calculation.In the process of cluster iterative upgrade,cluster management problems are gradually exposed.Cluster diversity and algorithm coupling make cluster management difficult.Nowadays,the management of clusters at home and abroad focuses on task scheduling and resource management.Clusters themselves are built based on the same nodes and use mature cluster management framework to form distributed computing clusters.The pulsar search computing cluster in the Key Laboratory of Information and Computing Science of Guizhou Province,the current main use independent of distributed computing and storage systems such as Craber calculation speed,its are calculated based on the existing node,designs the specific task scheduling program and database,realized the distributed computing method and the task scheduling,auxiliary pulsar search calculation.Aiming at the particularity of pulsar search computing cluster,a pulsar search computing service platform is designed to manage the cluster in a semi-dissociated management mode.On the one hand,the physical attributes of nodes are disassembled and managed in the way of logical computers.Pulsar search computing service platform serves as a factory,and the original management instructions for each attribute are analyzed and reconstructed and sent to the corresponding nodes.On the other hand,the microservice model is proposed to enable the original computing algorithm to split the structure into computing elements,and then form microservices through Docker and other outer packaging.At the same time the pulsar search service platform that integrates data display visual computing management and pulsar search results data retrieval,let original pulsar calculation process can intuitive experience and management,after collecting the user experience and feedback,the results of search and tagging is optimized integration,let the platform to manage from the calculation to the search process.Pulsar search computing service platform is based on Docker container technology to build a set of B/S architecture mode services to achieve the goals of rapid deployment,visual operation and management,simple operation and complex tasks,and pay attention to the integration of software and hardware and the scalability of update and iteration.Based on the cluster architecture concept,micro-service architecture model and platform architecture,this paper explains the key technologies and design ideas of pulsar search computing service platform,and shows the research and implementation of important functions.The pulsar computing and search platform can meet the application requirements of researchers for cluster node management,data and cluster visualization,and result retrieval.The integrated management makes deployment and expansion relatively easy,and provides a platform for the future development of pulsar computing cluster.