WLCG/HSF Workshop 2025

5 May 2025, 12:00 → 9 May 2025, 14:00 Europe/Paris

Auditorium P. Lehmann, Building 200 (IJCLab, Paris)

Welcome to the WLCG/HSF Workshop 2025 at IJCLab (Orsay, France), May 5-9 2025

WLCG/HSF Workshop @IJCLab

Monday 5 May

12:00 → 13:30 Plenary: Registration + Welcome Drink

13:30 → 14:35 Plenary: Opening

Conveners: Michel Jouvin (Université Paris-Saclay (FR)), Simone Campana (CERN)

13:30 Welcome to Paris

Speaker: Michel Jouvin (Université Paris-Saclay (FR))

13:45 Introduction - WLCG

Speakers: David Britton (University of Glasgow (GB)), Simone Campana (CERN), Tommaso Boccali

14:10 Introduction - HSF

Speakers: Eduardo Rodrigues (University of Liverpool (GB)), Paul James Laycock (Universite de Geneve (CH))

14:35 → 17:40 Plenary: AI and WLCG/HSF

14:35 AI & WLCG/HSF - scope of the session

Speakers: Alessandro Di Girolamo (CERN), Paul James Laycock (Universite de Geneve (CH))

14:40 AI and Facilities

Speaker: Ilija Vukotic (University of Chicago (US))

15:15 coffee/tea break

15:45 AI/ML for Physics use cases

Speaker: Tommaso Boccali

16:20 Heterogeneous resources and experiment software

Speaker: Paolo Calafiura (Lawrence Berkeley National Lab. (US))

16:55 Facilitiy view: Heterogenous resources deployment and usage

Speakers: Ben Jones (CERN), Rodney Walker (Ludwig Maximilians Universitat (DE))

17:20 Discussion

18:15 → 19:45 Social event: Welcome reception

Tuesday 6 May

08:30 → 09:00 Welcome coffee

09:00 → 12:30 HSF: Common software and Software Projects

09:00 ROOT: Taking stock of the Run 3 experience towards ROOT7

ROOT is a unified software package for the storage, processing, and analysis of scientific data: from its acquisition to the final visualization in the form of highly customizable, publication-ready plots. Successfully used by experiments and thousands of physicists, the ROOT Project is preparing its seventh release cycle, sustained by intense R&D activities.
In this contribution, after briefly reviewing the status of the project, we’ll focus on the results harvested along the R&D activities conducted so far and how those will shape the future of ROOT. The areas on which we’ll concentrate will mainly be three. Firstly, we discuss how hardware accelerators can be exploited by users with ROOT and how these new features are relevant at present and future Analysis Facilities. Then, we review the current development status of RNTuple and its adoption by experiments, illustrating both advancements in terms of performance but also usability, in C++ and Python. Finally, we’ll concentrate on the Python-C++ integration, at a low level, and the recent advancements in the interoperability of ROOT and Scientific Python, at the user-facing level.

Speaker: Danilo Piparo (CERN)

09:30 Latest Developments in RooFit and Plans

RooFit is a software package written in C++ for statistical data analysis that is part of ROOT. It is widely used in the High Energy Physics (HEP) community, with the most prominent users being the LHC collaborations. Recent RooFit development has focused on performance improvements and supporting new statistical analysis approaches to enable cutting-edge analyses, such as combined Higgs measurements with ATLAS or CMS. In this contribution, the development pillars that helped to achieve this goal are elaborated on. The first pillar is code optimization and refactoring to optimally use both CPU and GPU resources. Then, there is supporting Automatic Differentiation (AD) with Clad, a compiler plugin for Clang. Furthermore, RooFit now provides new Python interfaces to include ML models as likelihood surrogates, enabling Simulation-Based Inference (SBI). Finally, this contribution will also report on the development status of the Minuit2 library for numerical minimization since it is a key dependency of RooFit, and the two packages are developed hand-in-hand to implement performance-optimal statistical analysis workflows for HEP.

Speaker: Jonas Rembser (CERN)

10:00 Introduction to the Virtual Research Environment: an end-user perspective

One of the objectives of the EOSC (European Open Science Cloud) Future Project was to integrate diverse analysis workflows from Cosmology, Astrophysics and High Energy Physics in a common framework. This led to the inception of the Virtual Research Environment (VRE) at CERN, a prototype platform supporting the goals of Dark Matter and Extreme Universe Science Projects in compliance with FAIR (Findable, Accessible, Interoperable, Reusable) data policies. The goal of the project was to highlight the synergies between different dark matter communities and experiments, by producing new scientific results as well as by making the necessary data and software tools fully available. The VRE makes use of a common authentication and authorisation infrastructure (AAI), and shares the different experimental data (ATLAS, Fermi-LAT, CTA, Darkside, Km3Net, Virgo, LOFAR) in a reliable distributed storage infrastructure via the ESCAPE Data Lake. The entry point of such a platform (for an end-user) is a jupyterhub instance deployed on top of a scalable Kubernetes infrastructure, providing an interactive graphical interface for researchers to access, analyse and share data. The data access and browsability is enabled through API calls to the high level data management and storage orchestration software (Rucio). The VRE aims to streamline the development of end-to-end physics workflows, granting researchers access to an infrastructure that contains easy-to-use physics analysis workflow from different experiments. In this contribution, I will provide an overview of the VRE, highlight its use cases as an analyser for implementing and reproducing experimental analyses on a REANA cluster, and showcase the successful integration of an ATLAS experimental analysis workflow into the VRE platform.

Speaker: Sukanya Sinha (The University of Manchester (GB))

10:30 Coffee break

11:00 AdePT - Offloading electromagnetic showers in Geant4 simulations to GPU

The Geant4 simulation throughput of LHC experiments is limited by increasing detector complexity in the high-luminosity phase. As high-performance computing shifts toward heterogeneous architectures such as GPUs, GPU-accelerated particle transport simulations offer a potential way to improve performance. Currently, only electromagnetic showers can be offloaded to GPUs, making an efficient CPU–GPU workflow essential. In this contribution, we present state-of-the-art detector simulations for LHC experiments using GPUs, outline the outstanding challenges, and discuss future directions.

Speaker: Severin Diederichs (CERN)

11:20 Accelerating HEP detector simulations using G4HepEm

Geant4 based detector simulations make a significant contribution to the overall computing budget of the LHC experiments. The individual experiments have been investing considerable effort in making their simulations more and more efficient. These performance optimisations are now even more important in order to cope with the special computing challenges of the HL-LHC era.

G4HepEm is one of the R&D projects that have been launched with the goal of contributing to this effort. It provides an efficient simulation of the electromagnetic shower, tailored for HEP detector simulations, in the form of a Geant4 extension. A significant performance improvement (~20 %) of the ATLAS and CMS full detector simulations has been achieved recently after integrating G4HepEm into the ATLAS Athena and CMS-SW frameworks while preserving the accuracy of the results. The motivations, ideas and results obtained for ATLAS and CMS will be presented.

Speaker: Mihaly Novak (CERN)

09:00 → 10:30 WLCG: Token Migration Plan

Conveners: Alessandro Di Girolamo (CERN), Panos Paparrigopoulos (CERN)

10:30 → 11:00 Coffee break

11:00 → 12:30 WLCG: Facilities

Conveners: James Letts (Univ. of California San Diego (US)), Julia Andreeva (CERN)

12:30 → 14:15 Lunch

14:15 → 18:25 Plenary: Analysis at scale and analysis challenges

Conveners: Alessandra Forti (The University of Manchester (GB)), Dr Nicole Skidmore (University of Warwick)

14:15 Understanding the scale of HL-LHC physics analyses

The existing roadmaps and computing model plans from ATLAS and CMS for the HL-LHC area are primarily focused on the centralized aspect of computing: those steps that lead up to sets of files made available to physicists for analysis. The general approaches, resources used, and software frameworks for the area of “end-user physics analysis”, which starts from those files, are much less clearly defined, understood, or prescribed.

In order to better understand these aspects, IRIS-HEP, jointly with ATLAS AMG, CMS CAT and HSF DAAA, is designing a survey to capture the computational requirements from physics analysis use cases. This contribution will show first results from the survey and discuss the broader context of the effort. Following this survey, we aim to identify a set of physics analyses and define benchmark scenarios to extrapolate the concrete computing requirements to the HL-LHC era. The described workflows will provide more clarity about the role of analysis facilities and the kinds of services they should make available. It will allow for quantitative evaluation of analysis models and be a first step towards identifying what to do about analysis use cases that do not fit into the space set out by the benchmark examples.

Speakers: Alexander Held (University of Wisconsin Madison (US)), Oksana Shadura (University of Nebraska Lincoln (US))

14:45 Analysis at the HL-LHC: Data Delivery, ServiceX, and Addressing Our Analysis Challenges

As the HL-LHC era approaches, the scale and complexity of data present challenges for analysis workflows within ATLAS and other HL-LHC experiments. This contribution reports on recent developments in ServiceX, a cross experiment utility, and its role as a data delivery and transformation service within the analysis ecosystem. Designed to bridge the gap between centrally produced datasets and user-level analysis code, ServiceX now supports a broader range of input formats—including custom non-ROOT ATLAS data—and has seen significant improvements in reliability, scalability, and ease of use.

We highlight progress in several areas: improved bullet-proofing of the data transformation infrastructure to avoid errors at scale; new convenience utilities for introspecting datasets (e.g., listing available branches); and tighter integration with common ATLAS frameworks such as TopCPTools, enabling more standardized analysis environments. We also share results from recent scaling tests demonstrating performance at analysis-scale workloads, and discuss how ServiceX fits into the broader analysis software stack, helping to address long-standing challenges of reproducibility, data access, and user efficiency.

This talk will provide a view of the design of ServiceX, the technical and usability gains made in recent months, and what’s coming next as we prepare for the HL-LHC's analysis demands.

Speaker: Gordon Watts (University of Washington (US))

15:25 Coffee break

15:55 User experience

Speaker: Albert Gyorgy Borbely (University of Glasgow (GB))

16:10 Potential ROOT at scale talk?

16:25 ATLAS analysis facility contribution

16:40 CMS analysis facility contribution

Speaker: Oksana Shadura (University of Nebraska Lincoln (US))

16:55 LHCb analysis facility contribution

Speaker: Dr Nicole Skidmore (University of Warwick)

17:10 ALICE analysis facility contribution

17:25 Proposal for Interdisciplinary Analysis Facilities

Speaker: Thomas Hartmann (Deutsches Elektronen-Synchrotron (DE))

17:40 Discussion

Wednesday 7 May

08:30 → 09:00 Welcome coffee

09:00 → 12:30 HSF: Sustainable Software

09:00 The EVERSE/ESCAPE use cases: evaluating and improving the quality of HEP software

This contribution will present how the EVERSE project interfaces with the European Open Science Clusters (ENVRI-FAIR for environmental sciences, Life Sciences RI, ESCAPE for Particle physics and astrophysics, PaNOSC for Photon and neutron science and SSHOC for social sciences and humanities) through use cases of software packages or infrastructures that are in current used by researchers.

It is through these use cases that EVERSE draws best practices from the different communities, and where the various elements of the software excellence framework are tested and implemented prior to their release to the wider community. In this talk, we will describe the three ESCAPE use cases, highlight the elements that these use cases have contributed to EVERSE so far, and outline the expected improvements and pathway to obtain them by the end of the project in 2027.

Speakers: James Smith (The University of Manchester (GB)), Michael Sparks (BBC), Tobias Fitschen (The University of Manchester (GB))

09:20 The EVERSE Research Software Quality Toolkit

The Research Software Quality Toolkit (RSQKit - https://everse.software/RSQKit/), developed by the EVERSE project, lists curated best practices in improving the quality of your research software. It is intended for use by researchers, research software engineers, as well as those running research infrastructures involving software or involved in research software-related policy and funding.

These practices are informed by software excellence and quality in the context of research; with a focus on FAIR software, Open Research, community development and software engineering practices at different tiers of research software (analysis scripts, prototype tools and research software infrastructure). RSQKit links to tools and resources which support best practices. It includes software quality dimensions and links to indicators and tasks to guide the usage of the best practice, as well as links to training resources and existing guides and materials.

This contribution will introduce RSQKit and its aims and architecture. It is presented in a way to attract and consider feedback from researchers who code from the particle physics community and from WLCG infrastructure experts.

Speaker: Michael Sparks (BBC)

09:40 Green software in HEP: benchmarks and studies on MC generators

In this talk, we will describe the studies undertaken at the University of Manchester to estimate and improve the energy efficiency of computing hardware and software used by students and researchers.

The goal of these studies is to build an understanding of the environmental impact of paticle physics research focusing on two fronts:
1) the carbon cost of the hardware uses for high power computing hardware and the local computing cluster
2) the energy efficiency of data analysis software and machine learning models in “big data”-related scientific fields including as high-energy particle physics.

The focus of this contribution will be the energy efficiency of scientific software algorithms and MC generation packages, taking Herwig, ML data compression and top tagging algorithms as examples. We will discuss different tools and benchmarks and review their methodologies.

We will then describe our plans towards a lifecycle analysis for computing hardware, and work undergoing to estimate the power consumption of our local cluster more precisely.

Speakers: Luis Villar (University of Manchester), Tobias Fitschen (The University of Manchester (GB))

10:00 Scikit-HEP project news and future directions

Scikit-HEP is a community-driven and community-oriented project with the goal of providing an ecosystem for particle physics data analysis in Python fully integrated with the wider scientific Python ecosystem. The project provides many packages and a few “affiliated” packages for data analysis. It expands the typical Python data analysis tools for particle physicists, with packages spanning the spectrum from general scientific libraries for data manipulation to domain-specific libraries. An overview of where the project is will be presented. Future developments and matters of sustainability will be discussed.

Speaker: Eduardo Rodrigues (University of Liverpool (GB))

10:30 Coffee break

11:00 Marionette: Data Structure Description and Management for Heterogeneous Computing

Marionette is a header-only C++ library that was designed to allow the description of arbitrary data structures that can work across heterogeneous compute devices and on the host, providing complete interoperability and convenient interfaces with no impact on runtime performance. This is achieved by decoupling the description of the data to be held from the way in which data will be stored, which enables the generation of all memory allocations, transfers and deallocations at compile-time without requiring the end user to write any sort of boilerplate code, achieving the same performance as the equivalent hand-written alternatives. Furthermore, an expressive and intuitive object-oriented interface can be offered on both host and device(s), especially since arbitrary functions can be added to the interface of both the individual objects and the entire collection of them, without any runtime performance penalty as everything is resolved at compile-time. This user-extensible interface also means that the behaviour of pre-existing data structures can be replicated, allowing for immediate porting of pre-existing code with only minor adjustments to the data types. Furthermore, since the user can override and further specialize data transfers, gradual porting to the new data structures with minimal performance loss becomes possible as efficient ways to convert to and from pre-existing structures can be provided. With a focus on flexibility, customisability and extensibility without compromising expressivity, convenience or ease of use, Marionette is designed to offer a general solution for expressing data structures, catering to a wide variety of use cases and levels of expertise, with little to no runtime impact.

Speaker: Nuno Dos Santos Fernandes (Laboratory of Instrumentation and Experimental Particle Physics (PT))

11:20 Julia: Sustainability and Efficiency

There a number of studies of the general energy efficiency of different
programming languages, however relatively few look at HEP specific examples.
Here we present examples comparing energy efficiency of different jet
reconstruction codes in different languages: specifically C++, Julia and Python.
We also study the evolution of efficiency over recent releases of Julia and
Python.

We also discuss general aspects of sustainability of code and show how the Julia
language and ecosystem helps developers to write and maintain modular,
interoperable codes that reduce the code maintenance burden.

We show that Julia is an excellent language choice, combining outstanding energy
efficiency and human productivity, helping sustainability in all the most
meaningful senses.

Speaker: Graeme A Stewart (CERN)

09:00 → 12:30 WLCG: Technical Coordination

Conveners: Alessandro Di Girolamo (CERN), James Letts (Univ. of California San Diego (US))

10:30 Coffee break

12:30 → 14:15 Lunch

14:15 → 17:30 HSF: Training Working Group

14:15 The EVERSE training and recognition plan

The EVERSE project aims to collect, enhance and curate training resources aligned with domain-specific practices, create a long-term training activity supported by community services and platforms and establish a framework for recognizing Trainers and RSEs.

This contribution will describe how EVERSE plans to collect and provide training, guidance and education to researchers, software developers, and other stakeholders in the research community to help them understand the importance of software and code quality, as well as how to apply established best practices and standards for assessing, verifying, and improving the quality of their software and code.
We will describe the registry of training initiatives compiled so far, and how this registry will be maintained and presented through the TESS infrastructure. We will also outline recent activities in terms of recognition of software activities and roles.

Our aim is to get feedback from the community at the workshop and to highlight missing aspects in training and recognition that we can bring back to the project.

Speaker: Kenneth Brian Rioja (IT-FTI)

14:35 GWOSC and gravitational-wave data analysis training

Since their discovery in 2015, gravitational waves have become a hot topic in physics research.
Gravitational-wave data produced by the LVK Collaboration, formed by the LIGO, Virgo and KAGRA collaborations, become fully public after a grace period; combined with the relative simplicity of the data themselves (one time series of the main signal channel per each interferometer, plus some simple data quality information), this created a large community of professional scientists, students and even citizen scientists analyzing them.
The Gravitational-Wave Open Science Centre (gwosc.org), besides maintaining a site hosting the public data and relevant software, developed a large amount of training and teaching materials such as tutorials and documentation, and organises periodical GW Open Data Workshops, “crash courses” in GW data analysis. This contribution describes such materials and activities.

Speaker: Dr Stefano Bagnasco (Istituto Nazionale di Fisica Nucleare, Torino)

15:30 Coffee break

14:15 → 18:00 WLCG: Operations

Conveners: Julia Andreeva (CERN), Maarten Litmaath (CERN), Panos Paparrigopoulos (CERN)

14:15 Introduction to the OPS session

Speaker: Julia Andreeva (CERN)

14:20 AUDITOR. First results of the assessment by the WLCG sites

AUDITOR (Accounting Data Handling Toolbox for Opportunistic Resources) is a flexible and extensible accounting system designed to support a wide range of use cases and infrastructures. Its integration with APEL enables it to function as a generic component within the WLCG accounting infrastructure, tracking the usage of various types of site computing resources. Several WLCG sites have evaluated AUDITOR, and the outcomes of these assessments will be reviewed and discussed.

Speakers: Alessandro Pascolini (Universita e INFN, Bologna (IT)), Alexander Raphael Kleinemuhl (Bergische Universitaet Wuppertal (DE)), Maria Alandes Pradillo (CERN), Max Fischer (Karlsruhe Institute of Technology), Michael Boehler (Deutsches Elektronen-Synchrotron (DE)), Michael Boehler (University of Freiburg (DE)), Thomas Hartmann (Deutsches Elektronen-Synchrotron (DE))

15:50 Coffee break

16:10 Accounting of jobs submitted with tokens and changes required in the WLCG accountig infrastructure.

16:25 GGUS: past, present, future

17:10 Job Allocation and Handling, status report from the WG

Speaker: Antonio Perez-Calero Yzquierdo (Centro de Investigaciones Energéticas Medioambientales y Tecnológicas)

Thursday 8 May

08:30 → 09:00 Welcome coffee

09:00 → 12:30 Plenary: Environmental Sustainability

Conveners: Caterina Doglioni (The University of Manchester (GB)), David Britton (University of Glasgow (GB))

10:30 Coffee break

12:30 → 14:15 Lunch

14:15 → 17:30 HSF: BOF (Birds Of a Feather)

15:30 Coffee break

14:15 → 17:30 WLCG: DOMA - Working Groups and Data Challenges

Conveners: Johannes Elmsheuser (Brookhaven National Laboratory (US)), Katy Ellis (Science and Technology Facilities Council STFC (GB))

14:15 Introduction and Status of DOMA

Speakers: Johannes Elmsheuser (Brookhaven National Laboratory (US)), Katy Ellis (Science and Technology Facilities Council STFC (GB))

14:35 Rucio in DOMA

Speaker: Martin Barisits (CERN)

14:55 FTS status and plans towards DC27

Speaker: Mihai Patrascoiu (CERN)

15:15 Coffee break

15:45 Status of Monitoring of DOMA components

Speaker: Borja Garrido Bear (CERN)

16:05 Status of DOMA BDT activity

Speakers: Mihai Patrascoiu (CERN), Petr Vokac (Czech Technical University in Prague (CZ))

16:30 Mini data challenges

Speakers: Johannes Elmsheuser (Brookhaven National Laboratory (US)), Katy Ellis (Science and Technology Facilities Council STFC (GB))

15:00 → 17:00 WLCG: Collaboration Board

20:00 → 22:00 Social Dinner

in La Mère Catherine, https://lamerecatherine.com/

Friday 9 May

08:30 → 09:00 Welcome coffee

09:00 → 12:00 Plenary: Closing

10:30 Coffee break

12:00 → 14:00 Lunch

Be sure to say that you'll attend the lunch during the registration