Keyword: GPU
Paper Title Other Keywords Page
TU1BCO01 A Workflow for Training and Deploying Machine Learning Models to EPICS controls, EPICS, framework, software 244
 
  • M.F. Leputa, K.R.L. Baker, M. Romanovschi
    STFC/RAL/ISIS, Chilton, Didcot, Oxon, United Kingdom
 
  The transition to EPICS as the control system for the ISIS Neutron and Muon Source accelerators is an opportunity to more easily integrate machine learning into operations. But developing high quality machine learning (ML) models is insufficient. Integration into critical operations requires good development practices to ensure stability and reliability during deployment and to allow robust and easy maintenance. For these reasons we implemented a workflow for training and deploying models that utilize off-the-shelf, industry-standard tools such as MLflow. Our experience of how adoption of these tools can make developer’s lives easier during the training phase of a project is discussed. We describe how these tools may be used in an automated deployment pipeline to allow the ML model to interact with our EPICS ecosystem through Python-based IOCs within a containerized environment. This reduces the developer effort required to produce GUIs to interact with the models within the ISIS Main Control Room as tools familiar to operators, such as Phoebus, may be used.  
slides icon Slides TU1BCO01 [3.370 MB]  
DOI • reference for this paper ※ doi:10.18429/JACoW-ICALEPCS2023-TU1BCO01  
About • Received ※ 05 October 2023 — Accepted ※ 12 October 2023 — Issued ※ 19 October 2023  
Cite • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THMBCMO31 LImA2: Edge Distributed Acquisition and Processing Framework for High Performance 2D Detectors detector, controls, SRF, experiment 1269
 
  • S. Debionne, L. Claustre, P. Fajardo, A. Götz, A. Homs Puron, J. Kieffer, R. Ponsard
    ESRF, Grenoble, France
 
  LImA* is a framework born at the ESRF for 2D Data Acquisition (DAQ), basic Online Data Analysis (ODA) and processing with high-throughput detectors. While in production for 15 years in several synchrotron facilities, the ever-increasing detector frame rates make more and more difficult performing DAQ & ODA tasks on a single computer**. LImA2 is designed to scale horizontally, using multiple hosts for DAQ & ODA. This enables more advanced strategies for data feature extraction while keeping a low latency. LImA2 separates three functional blocks: detector control, image acquisition, and data processing. A control process configures the detector, while one or more receiver processes perform the DAQ and ODA, like the generation of fast feedback signals. The detectors currently supported in LImA2 are the PSI/Jungfrau, the ESRF/Smartpix and the Dectris/Eiger2. The former performs pixel assembly and intensity correction in GPU; the second exploits RoCE capabilities; and the latter features dual threshold, multi-band images. Raw data rates up to 8 GByte/s can be handled by a single computer, scalable if necessary. In addition to a classic processing, advanced pipelines are also implemented. A Serial-MX/pyFAI*** pipeline extracts diffraction peaks in GPU in order to filter low quality data. NVIDIA GPUDirect is used by a third pipeline providing 2D processing with remarkable low latency. IBM Power9 optimizations like the NX GZIP compression and the PCI-e multi-host extension are exploited.
* LIMA - https://accelconf.web.cern.ch/ICALEPCS2013/papers/frcoaab08.pdf
** Jungfraujoch - https://doi.org/10.1107/S1600577522010268
*** pyFAI - https://doi.org/10.1107/S1600576715004306
 
slides icon Slides THMBCMO31 [0.572 MB]  
poster icon Poster THMBCMO31 [14.959 MB]  
DOI • reference for this paper ※ doi:10.18429/JACoW-ICALEPCS2023-THMBCMO31  
About • Received ※ 06 October 2023 — Revised ※ 08 October 2023 — Accepted ※ 11 December 2023 — Issued ※ 13 December 2023
Cite • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)  
 
THPDP079 Integration of Bespoke Daq Software with Tango Controls in the SKAO Software Framework: From Problems to Progress TANGO, controls, data-acquisition, software 1533
 
  • A.J. Clemens
    OSL, St Ives, Cambridgeshire, United Kingdom
  • D. Devereux
    CSIRO, Clayton, Australia
  • D. Devereux
    SKAO, Macclesfield, United Kingdom
  • A. Magro
    ISSA, Msida, Malta
 
  The Square Kilometre Array Observatory (SKAO) project is an international effort to build two radio interferometers in South Africa and Australia to form one Observatory monitored and controlled from the global headquarters in the United Kingdom at Jodrell Bank. The Monitoring, Control and Calibration System (MCCS) is the "front-end" management software for the Low telescope which provides monitoring and control capabilities as well as implementing calibration processes and providing complex diagnostics support. Once completed the Low telescope will boast over 130, 000 individual log-periodic antennas and so the scale of the data generated will be huge. It is estimated that an average of 8 terabits per second of data will be transferred from the SKAO telescopes in both countries to Central Processing Facilities (CPFs) located at the telescope sites. In order to keep pace with this magnitude of data production an equally impressive data acquisition (DAQ) system is required. This paper outlines the challenges encountered and solutions adopted whilst incorporating a bespoke DAQ library within the SKAO’s Kubernetes-Tango ecosystem in the MCCS subsystem in order to allow high speed data capture whilst maintaining a consistent deployment experience.  
poster icon Poster THPDP079 [0.981 MB]  
DOI • reference for this paper ※ doi:10.18429/JACoW-ICALEPCS2023-THPDP079  
About • Received ※ 02 October 2023 — Accepted ※ 08 December 2023 — Issued ※ 19 December 2023  
Cite • reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)