ICALEPCS2021 - List of Authors (Gorbonosov, R.)

Paper	Title	Page
MOBL03	Machine Learning Platform: Deploying and Managing Models in the CERN Control System	50
	J.-B. de Martel, R. Gorbonosov, N. Madysa CERN, Geneva, Switzerland
	Recent advances make machine learning (ML) a powerful tool to cope with the inherent complexity of accelerators, the large number of degrees of freedom and continuously drifting machine characteristics. A diverse set of ML ecosystems, frameworks and tools are already being used at CERN for a variety of use cases such as optimization, anomaly detection and forecasting. We have adopted a unified approach to model storage, versioning and deployment which accommodates this diversity, and we apply software engineering best practices to achieve the reproducibility needed in the mission-critical context of particle accelerator controls. This paper describes CERN Machine Learning Platform - our central platform for storing, versioning and deploying ML models in the CERN Control Center. We present a unified solution which allows users to create, update and deploy models with minimal effort, without constraining their workflow or restricting their choice of tools. It also provides tooling to automate seamless model updates as the machine characteristics evolve. Moreover, the system allows model developers to focus on domain-specific development by abstracting infrastructural concerns.
	Slides MOBL03 [0.687 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2021-MOBL03
About •	Received ※ 07 October 2021 Accepted ※ 16 November 2021 Issue date ※ 07 February 2022
Cite •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

TUPV033	Distributed Transactions in CERN’s Accelerator Control System	468
	F. Hoguin, S. Deghaye, R. Gorbonosov, J. Lauener, P. Mantion CERN, Geneva, Switzerland
	Devices in CERN’s accelerator complex are controlled through individual requests, which change settings atomically on single Devices. Individual Devices are therefore controlled transactionally. Operators often need to apply a set of changes which affect multiple devices. This is achieved by sending requests in parallel, in a minimum amount of time. However, if a request fails, the Control system ends up in an undefined state, and recovering is a time-consuming task. Furthermore, the lack of synchronisation in the application of new settings may lead to the degradation of the beam characteristics, because of settings being partially applied. To address these issues, a protocol was developed to support distributed transactions and commit synchronisation in the CERN Control system, which was then implemented in CERN’s real-time frameworks. We describe what this protocol intends to solve and its limitations. We also delve into the real-time framework implementation and how developers can benefit from the 2-phase commit to leverage hardware features such as double buffering, and from the commit synchronisation allowing settings to be changed safely while the accelerator is operational.
	Poster TUPV033 [0.869 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2021-TUPV033
About •	Received ※ 09 October 2021 Revised ※ 18 October 2021 Accepted ※ 20 November 2021 Issue date ※ 22 January 2022
Cite •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

TUPV047	Controlling the CERN Experimental Area Beams	509
	B. Rae, V. Baggiolini, D. Banerjee, J. Bernhard, M. Brugger, N. Charitonidis, M. Gabriel, A. Gerbershagen, R. Gorbonosov, M. Hrabia, M. Peryt, C. Roderick, G. Romagnoli CERN, Geneva, Switzerland L. Gatignon Lancaster University, Lancaster, United Kingdom
	The CERN fixed target experimental areas are comprised of more than 8km of beam line with around 800 devices used to control and measure the beam. Each year more than 140 groups of users come to perform experiments in these areas, with a need to access the data from these devices. The software to allow this therefore has to be simple, robust, and be able to control and read out all types of beam devices. This contribution describes the functionality of the beamline control system, CESAR, and its evolution. This includes all the features that can be used by the beamline physicists, operators, and device experts that work in the experimental areas. It also underlines the flexibility that the software provides to the experimental users for control of their beam line during data taking, allowing them to manage this in a very easy and independent way. This contribution also covers the on-going work of providing MAD-X support to CESAR to achieve an easier way of developing and integrating beam optics. An overview of the on-going software migration of the Experimental Areas is also given.
	Poster TUPV047 [1.262 MB]
DOI •	reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2021-TUPV047
About •	Received ※ 11 October 2021 Revised ※ 21 October 2021 Accepted ※ 21 December 2021 Issue date ※ 18 January 2022
Cite •	reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

Paper

Title

Page

Machine Learning Platform: Deploying and Managing Models in the CERN Control System

50

J.-B. de Martel, R. Gorbonosov, N. Madysa
CERN, Geneva, Switzerland

Recent advances make machine learning (ML) a powerful tool to cope with the inherent complexity of accelerators, the large number of degrees of freedom and continuously drifting machine characteristics. A diverse set of ML ecosystems, frameworks and tools are already being used at CERN for a variety of use cases such as optimization, anomaly detection and forecasting. We have adopted a unified approach to model storage, versioning and deployment which accommodates this diversity, and we apply software engineering best practices to achieve the reproducibility needed in the mission-critical context of particle accelerator controls. This paper describes CERN Machine Learning Platform - our central platform for storing, versioning and deploying ML models in the CERN Control Center. We present a unified solution which allows users to create, update and deploy models with minimal effort, without constraining their workflow or restricting their choice of tools. It also provides tooling to automate seamless model updates as the machine characteristics evolve. Moreover, the system allows model developers to focus on domain-specific development by abstracting infrastructural concerns.

Slides MOBL03 [0.687 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2021-MOBL03

About •

Received ※ 07 October 2021 Accepted ※ 16 November 2021 Issue date ※ 07 February 2022

Cite •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

TUPV033

Distributed Transactions in CERN’s Accelerator Control System

468

F. Hoguin, S. Deghaye, R. Gorbonosov, J. Lauener, P. Mantion
CERN, Geneva, Switzerland

Devices in CERN’s accelerator complex are controlled through individual requests, which change settings atomically on single Devices. Individual Devices are therefore controlled transactionally. Operators often need to apply a set of changes which affect multiple devices. This is achieved by sending requests in parallel, in a minimum amount of time. However, if a request fails, the Control system ends up in an undefined state, and recovering is a time-consuming task. Furthermore, the lack of synchronisation in the application of new settings may lead to the degradation of the beam characteristics, because of settings being partially applied. To address these issues, a protocol was developed to support distributed transactions and commit synchronisation in the CERN Control system, which was then implemented in CERN’s real-time frameworks. We describe what this protocol intends to solve and its limitations. We also delve into the real-time framework implementation and how developers can benefit from the 2-phase commit to leverage hardware features such as double buffering, and from the commit synchronisation allowing settings to be changed safely while the accelerator is operational.

Poster TUPV033 [0.869 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2021-TUPV033

About •

Received ※ 09 October 2021 Revised ※ 18 October 2021 Accepted ※ 20 November 2021 Issue date ※ 22 January 2022

Cite •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)

TUPV047

Controlling the CERN Experimental Area Beams

509

B. Rae, V. Baggiolini, D. Banerjee, J. Bernhard, M. Brugger, N. Charitonidis, M. Gabriel, A. Gerbershagen, R. Gorbonosov, M. Hrabia, M. Peryt, C. Roderick, G. Romagnoli
CERN, Geneva, Switzerland
L. Gatignon
Lancaster University, Lancaster, United Kingdom

The CERN fixed target experimental areas are comprised of more than 8km of beam line with around 800 devices used to control and measure the beam. Each year more than 140 groups of users come to perform experiments in these areas, with a need to access the data from these devices. The software to allow this therefore has to be simple, robust, and be able to control and read out all types of beam devices. This contribution describes the functionality of the beamline control system, CESAR, and its evolution. This includes all the features that can be used by the beamline physicists, operators, and device experts that work in the experimental areas. It also underlines the flexibility that the software provides to the experimental users for control of their beam line during data taking, allowing them to manage this in a very easy and independent way. This contribution also covers the on-going work of providing MAD-X support to CESAR to achieve an easier way of developing and integrating beam optics. An overview of the on-going software migration of the Experimental Areas is also given.

Poster TUPV047 [1.262 MB]

DOI •

reference for this paper ※ https://doi.org/10.18429/JACoW-ICALEPCS2021-TUPV047

About •

Received ※ 11 October 2021 Revised ※ 21 October 2021 Accepted ※ 21 December 2021 Issue date ※ 18 January 2022

Cite •

reference for this paper using ※ BibTeX, ※ LaTeX, ※ Text/Word, ※ RIS, ※ EndNote (xml)