Research on data strorage and reading method of EAST poloidal field power supply

The plasma discharge duration of EAST (Experimental Advanced Super—conducting Tokamak) is designed to be 1000s, the time has exceeded 100s at present. As the core subsystem, poloidal filed power supply(PF) has nearly 200 signals to be stored. For every discharge, the amount of data is very large, the ordinary database can’t meet the requirement of long pulse data storage. This paper uses HDF5 (Hierarchical Data Format) to store the long pulse data, HDF5 is a file format that can store different types of images and data. Using C interface of HDF5 to design a dataset with compound datatype, to realize real-time writing and reading, the primary discharge time is segmented and the data is wrote and read as chunk. Visualization tool Hdfview is used to view the dataset, and Qt is used to read HDF5 file and display waveform. It proves the correctness of compound datatype and chunk dataset, and the feasibility of writing and reading as chunk. Simple time test proves that writing and reading as chunk are efficient.

EAST（实验先进超导托卡马克，Experimental Advanced Superconducting Tokamak）的等离子体放电时长设计指标为1000秒，目前已实现超过100秒的放电运行。极向场电源（PF，poloidal field power supply）作为其核心子系统，需存储近200路监测信号。单次放电产生的数据体量极大，普通数据库无法满足长脉冲数据的存储需求。本文采用HDF5（分层数据格式，Hierarchical Data Format）存储长脉冲数据，HDF5是一种可兼容多种图像与数据类型的文件格式。依托HDF5的C接口设计复合数据类型数据集，实现数据的实时读写；将单次放电的总时长进行分段，以分块（chunk）形式完成数据的读写操作。使用可视化工具Hdfview查看该数据集，并基于Qt平台读取HDF5文件并绘制波形曲线。实验验证了复合数据类型与分块数据集的正确性，以及分块读写方案的可行性。简易时序测试表明，该分块读写方式具备较高的运行效率。