Implementation and Performance Analysis of the Archive System for the Taiwan Photon Source Project

Abstract

The EPICS framework was chosen as control system infrastructure for the new project of 3 GeV synchrotron light source (Taiwan Photon Source, TPS). Various control system prototypes of TPS with the EPICS mechanism have been gradually built and tested. One EPICS data archive system is necessary to be developed to record various machine parameters and status information for long time logging. An archive engine takes PV data from EPICS IOCs via channel access, and stores them in the data storage. The PostgreSQL RDB (Relational Database) was adopted as the data storage for the TPS archive system. Both the historic PVs (Process Variable) data and the archive engine configuration are saved into the same RDB. The archived data can be retrieved in the graphical user interface of data browser. Taking the performance and redundancy into considerations, the storage servers and RDB table structures are tuned relatively. The primary factor most common used to judge system performance is the mean response time. We propose the queueing model method to analyze the performance of a network-based archive system, and simplify the architecture as the queueing model in serial and parallel to derive the equations of mean response time. Then we apply the proposed method to estimate the approximate response time of the actual network-based system which constructed in serial and in parallel. The efforts are described at this paper.