FRCB —  Contributed Oral: Data Acquisitions 2 and Hardware Technologies 2   (19-Oct-18   10:50—11:50)
Chair: P. Duval, DESY, Hamburg, Germany
Paper Title Page
FRCB1 Ultra Fast Data Acquisition in ELI Beamlines -1
  • P. Bastl
    Institute of Physics of the ASCR, Prague, Czech Republic
  • V. Gaman, O. Janda, P. Pivonka, B. Plötzeneder, J. Sys, J. Trdlicka
    ELI-BEAMS, Prague, Czech Republic
  The ELI Beamlines facility is a Petawatt laser facility in the final construction and commissioning phase in Czech Republic. In fully operation phase, four lasers will be used to control beamlines in six experimental halls. In this paper we describe Ultra fast and distributed data acquisition system as was defined in ELI Beamlines. The data acquisition system is divided into two levels: central and local level. The central level data acquisition system defines a special Tier 0 RAM buffer. This buffer is based on special multi node data acquisition server which shares memory of all its nodes into one continuous space over low latency network technologies (Mellanox Infinband/Intel OmniPath). The main role of the Tier 0 buffer is to acquire first bunch and provide load balancing of incoming data. These data comes from many sources distributed along the experimental technologies. The local data acquisition system is then responsible for connection of local detectors to central data acquisition server through ROCE interface. The connection is done directly when supported or indirectly using local data acquisition computers (for PCIe etc.).  
FRCB2 Design and Construction of the Data Warehouse Based on Hadoop Ecosystem at HLS-II -1
  • Y. Song, X. Chen, C. Li, G. Liu, J.G. Wang, K. Xuan
    USTC/NSRL, Hefei, Anhui, People's Republic of China
  Funding: Work supported by National Natural Science Foundation of China (No.11375186)
A data warehouse based on Hadoop ecosystem is designed and constructed for Hefei Light Source II (HLS-II). The ETL program based on Spark migrates data to HDFS from RDB Channel Archiver and the EPICS Archiver Appliance continuously and store them in Parquet format. The distributed data analysis engine based on Impala greatly improves the performance of data retrieval and reduces the response time of queries. In this paper, we will describe our efforts and experience to use various open sources software and tools to effectively manage the big data. We will also report the plans on this data warehouse in the future.
Novel Concept of Off-detector Electronics Based on Machine Learning for High Energy Physics  
  • W. Wang. Wang, B.M. Balzer, M. Brunet, M. Caselle, L. Rota, M. Weber
    KIT, Karlsruhe, Germany
  Due to increase of complexity of HEP detectors operating in rush radiation environment, detector calibration operations becomes a challenge. Moreover, growing volumes of available data, traditional calibration approach risk being weak. Therefore, new methods that can automatically analyses large complex data and deliver quick accurate results, without human intervention, are necessary in modern detectors. Embedded machine learning algorithms on FPGAs enables new methods for performing high-granularity Data Quality Monitoring (DQM) and real-time detector calibration, ensuring optimal data quality for further offline analysis. Combining FPGA with multiprocessor system-on-chip (MPSoC) opens "new ideas" in integration of DQM and slow-control for off-detector electronics. In this context, a novel readout electronics based on ZYNQ Ultrascale+ MPSoC has been designed. To handle the huge data volume, the ZYNQ board is equipped with FireFly optical data links. In this contribution, a novel high-granularity Data Quality Monitoring (DQM) architecture with real-time detector calibration based on embedded machine learning is presented.  
FRCB4 The Application for Fault Diagnosis and Prediction of Power Supply Control Device on BEPCII -1
  • J. Liu, D. Wang, J.C. Wang, X.L. Wang
    IHEP, Beijing, People's Republic of China
  With the widely adoption of complex electronic devices and microcircuits in accelerator system, the probability of system failure and functional failure will be enlarged. For example, the fault of the magnet power supply front-end electronics devices may cause accelerator energy instability and even lead to beam loss. Therefore, it is very necessary to diagnose and locate the device fault accurately and rapidly, that will induce the high cost of the accelerator operation. Faults diagnosis and prediction can not only improve the safety and reliability of the equipment, but also effectively reduce the equipment's cycle costing. We applied the FMECA and testability modeling method for the PSI device, which using in BEPCII power supply control system, and evaluated the remaining life of the PSI under certain temperature and humidity condition based on the reliability model and accelerated life test.