Device-Edge-Cloud Intelligent Collaboration framEwork

Informazioni relative al progetto

DECICE

ID dell’accordo di sovvenzione: 101092582

DOI

10.3030/101092582

Progetto chiuso

Data della firma CE 23 Novembre 2022

Data di avvio 1 Dicembre 2022

Data di completamento 30 Novembre 2025

Finanziato da

• Digital, Industry and Space

Costo totale

€ 5 627 250,00

Contributo UE

€ 5 627 249,50

Coordinato da

GEORG-AUGUST-UNIVERSITAT GOTTINGEN STIFTUNG OFFENTLICHEN RECHTS
Germany

CORDIS fornisce collegamenti ai risultati finali pubblici e alle pubblicazioni dei progetti ORIZZONTE.

I link ai risultati e alle pubblicazioni dei progetti del 7° PQ, così come i link ad alcuni tipi di risultati specifici come dataset e software, sono recuperati dinamicamente da .OpenAIRE .

Risultati finali

Documents, reports (10)

Implementation Report of CI/CD Environment

Technical description of the selected CI/CD environment including the chosen code quality and security tools. This further includes the description of the configuration of quality gates preventing poor quality or insecure code from being built.

Use Case Requirements

A list of the selected use cases and a detailed inventory of requirements for each application use case of the DECICE framework that covers hardware specifications (e.g. networking, storage, computinge) and performance demands.

Specification of the Optimization Scope

A detailed technical description of the system characteristics and metrics necessary for the scheduler and the supported user-requirements for job/workload.

AI-Scheduler Prototypes for Storage and Compute

Individual prototypes of the two AI-schedulers for storage and compute are implemented and are ready for use in the synthetic training environment.

Dissemination & Communication Plan

This deliverable will be focused on informing the broad audience on the project’s goals and activities, and raising awareness on its results. To achieve this, DECICE will lay out the target stakeholders as well as the means to reach them.  This will be updated in M18 for adjusting the plan to the requirements of the project and new opportunities and challenges. The plan will lay out the various activities which will seek to encourage stakeholders to use and uptake the AI-techniques and models developed by the project for further research or commercial activities.

Integration of Monitoring Framework

This deliverable contains the software components, technical descriptions and documentation for the monitoring framework.

Project Handbook

This document provides an overview of management and administrative procedures that are in place to guarantee the quality of project results and efficient teamwork. Key aspects of these procedures and standards are discussed and agreed upon during the kick-off general assembly. This includes a description of the collaborative environment, how to organize material, a FAQ for typical questions of project partners, and plans for quality assurance, technical integration, risk management, communication and dissemination. For example, the QA and risk management plan defines guidelines, procedures, and checklists to be followed for project specific documentation, criteria for the reviews, communication, innovation potential, progress and impact indicators, and contingency plans. The plan also includes assigned deliverable and milestone review teams and documents the potential and emerged risks. Throughout the project, the project handbook will be constantly updated.

Development Environment Specification

A detailed technical specification of the necessary hardware and software to create a development environment for the DECICE framework as well as the definition of performance metrics for evaluation.

Online and Media Presence

This deliverable will present information on the development of the project website, the utilized social media channels and project newsletter.

Data Management Plan

The first version of the DMP is based on the Horizon Europe guidelines. It ensures that created data and research results are findable, accessible, interoperable and reusable (FAIR). It covers the handling of data during and after the project: the generated, collected and processed data, the procedures and standards applied, the restrictions on data, and the data curation and preservation. The DMP will be periodically updated similarly to the project handbook.

Demonstrators, pilots, prototypes (1)

Project Development Environment Deployed for Phase 1 and 2

A project development environment is deployed for Phase 1 (one site) and Phase 2 (2 sites) with access for all project members to enable performance evaluations.

Other (2)

Synthetic Test Environment

This deliverable contains the code and documentation for the synthetic test environment as defined in T3.5 that allows the training of the ML-models from WP2. Slight adaptations may be necessary after the initial version while the overall system architecture is not yet finalized. It also contains a description of the architecture.

Digital Twin

Core of this deliverable is the implementation of the digital twin and the corresponding interfaces, including the integration of the Telemetry API and the interface to the AI-schedulers.

Pubblicazioni

Peer reviewed articles (7)

GRAAFE: GRaph Anomaly Anticipation Framework for Exascale HPC systems

Autori: Martin Molan, Mohsen Seyedkazemi Ardebili, Junaid Ahmed Khan, Francesco Beneventi, Daniele Cesarini, Andrea Borghesi, Andrea Bartolini
Pubblicato in: Future Generation Computer Systems, Numero 160, 2024, ISSN 0167-739X
Editore: Elsevier BV
DOI: 10.1016/J.FUTURE.2024.06.032

The REGALE Library: A DDS Interoperability Layer for the HPC PowerStack

Autori: Giacomo Madella, Federico Tesser, Lluis Alonso, Julita Corbalan, Daniele Cesarini, Andrea Bartolini
Pubblicato in: Journal of Low Power Electronics and Applications, Numero 15, 2025, ISSN 2079-9268
Editore: MDPI AG
DOI: 10.3390/JLPEA15010010

Performance Characterization of Hardware/Software Communication Interfaces in End-to-End Power Management Solutions of High-Performance Computing Processors

Autori: Antonio del Vecchio, Alessandro Ottaviano, Giovanni Bambini, Andrea Acquaviva, Andrea Bartolini
Pubblicato in: Energies, Numero 17, 2025, ISSN 1996-1073
Editore: MDPI AG
DOI: 10.3390/EN17225778

Directly-trained Spiking Neural Networks for Deep Reinforcement Learning: Energy efficient implementation of event-based obstacle avoidance on a neuromorphic accelerator

Autori: Luca Zanatta, Alfio Di Mauro, Francesco Barchi, Andrea Bartolini, Luca Benini, Andrea Acquaviva
Pubblicato in: Neurocomputing, Numero 562, 2025, ISSN 0925-2312
Editore: Elsevier BV
DOI: 10.1016/J.NEUCOM.2023.126885

M100 ExaData: a data collection campaign on the CINECA’s Marconi100 Tier-0 supercomputer

Autori: Borghesi, Andrea; Di Santi, Carmine; Molan, Martin; Seyedkazemi Ardebili, Mohsen; Mauri, Alessio; Guarrasi, Massimiliano; Galetti, Daniela; Cestari, Mirko; Barchi, Francesco; BENINI, LUCA; Beneventi, Francesco; Bartolini, Andrea
Pubblicato in: Scientific Data, Vol 10, Iss 1, Pp 1-10 (2023), 2023, ISSN 2052-4463
Editore: Nature
DOI: 10.3929/ethz-b-000614369

Fair and efficient resource allocation via vehicle-edge cooperation in 5G-V2X networks

Autori: Muhammed Nur Avcil, Mujdat Soyturk, Burak Kantarci
Pubblicato in: Vehicular Communications, Numero 48, 2024, ISSN 2214-2096
Editore: Elsevier BV
DOI: 10.1016/J.VEHCOM.2024.100773

HazardNet: A thermal hazard prediction framework for datacenters

Autori: Seyedkazemi Ardebili, Mohsen, Andrea Acquaviva, Luca Benini, and Andrea Bartolini.
Pubblicato in: Future Generation Computer Systems, ISSN 1872-7115
Editore: ScienceDirect
DOI: 10.1016/J.FUTURE.2024.01.031

Conference proceedings (9)

PM100: A Job Power Consumption Dataset of a Large-scale Production HPC System

Autori: Francesco Antici, Mohsen Seyedkazemi Ardebili, Andrea Bartolini, Zeynep Kiziltan
Pubblicato in: Proceedings of the SC '23 Workshops of the International Conference on High Performance Computing, Network, Storage, and Analysis, 2025
Editore: ACM
DOI: 10.1145/3624062.3624263

DECICE

Autori: Julian Martin Kunkel; Christian Boehme; Jonathan Decker; Fabrizio Magugliani; Dirk Pleiter; Bastian Koller; Karthee Sivalingam; Sabri Pllana; Alexander Nikolov; Mujdat Soyturk; Christian Racca; Andrea Bartolini; Adrian Tate; Berkay Yaman
Pubblicato in: Crossref, 2023, ISBN 9798400701405
Editore: Association for Computing Machinery
DOI: 10.48550/arxiv.2305.02697

SpikeStream: Accelerating Spiking Neural Network Inference on RISC-V Clusters with Sparse Computation Extensions

Autori: Simone Manoni, Paul Scheffler, Luca Zanatta, Andrea Acquaviva, Luca Benini, Andrea Bartolini
Pubblicato in: 2025 Design, Automation & Test in Europe Conference (DATE), 2025
Editore: IEEE
DOI: 10.23919/DATE64628.2025.10992749

TinyLid: a RISC-V accelerated Neural Network For LiDAR Contaminant Classification in Autonomous Vehicle

Autori: Grafika Jati, Martin Molan, Francesco Barchi, Andrea Bartolini, Andrea Acquaviva
Pubblicato in: Proceedings of the 21st ACM International Conference on Computing Frontiers, 2025
Editore: ACM
DOI: 10.1145/3649153.3649201

Comparing Fault-tolerance in Kubernetes and Slurm in HPC Infrastructure

Autori: Mirac Aydin, Michael Bidollahkhani, and Julian M. Kunkel
Pubblicato in: ISSN 2308-4499
Editore: ADVCOMP 2024

Graph Neural Networks for Anomaly Anticipation in HPC Systems

Autori: Martin Molan, Junaid Ahmed Khan, Andrea Borghesi, Andrea Bartolini
Pubblicato in: Companion of the 2023 ACM/SPEC International Conference on Performance Engineering, 2025
Editore: ACM
DOI: 10.1145/3578245.3585335

Bio-Inspired Drone Control: A Reinforcement Learning-Trained Spiking Neural Networks for Agile Navigation in Dynamic Environment

Autori: Yin-Ching Lee, Sebastiano Mengozzi, Luca Zanatta, Andrea Bartolini, Andrea Acquaviva, Francesco Barchi
Pubblicato in: 2025 IEEE International Conference on Omni-layer Intelligent Systems (COINS), 2025
Editore: IEEE
DOI: 10.1109/COINS65080.2025.11125776

HOSHMAND: Accelerated AI-Driven Scheduler Emulating Conventional Task Distribution Techniques for Cloud Workloads

Autori: Michael Bidollahkhani, Aasish Kumar Sharma and Julian Martin Kunkel
Pubblicato in: Numero CFP24061-ART, ISSN 2836-3795
Editore: IEEE COMPSAC

Towards Nano-Drones Agile Flight Using Deep Reinforcement Learning

Autori: Sebastiano Mengozzi, Luca Zanatta, Francesco Barchi, Andrea Bartolini, Andrea Acquaviva
Pubblicato in: 2024 IEEE International Conference on Omni-layer Intelligent Systems (COINS), 2024
Editore: IEEE
DOI: 10.1109/COINS61597.2024.10622558

Other (4)

Monte Cimone v2: Down the Road of RISC-V High-Performance Computers

Editore: Emanuele Venieri, Simone Manoni, Gabriele Ceccolini, Giacomo Madella, Federico Ficarelli, Daniele Gregori, Daniele Cesarini, Luca Benini, Andrea Bartolini

https://arxiv.org/html/2504.01972

Autori: Umberto Laghi, Simone Manoni, Emanuele Parisi, Andrea Bartolini
Pubblicato in: 2025
Editore: arxiv.org

MCBound: An Online Framework to Characterize and Classify Memory/Compute-bound HPC Jobs

Autori: Antici, Francesco ; Bartolini, Andrea ; Kiziltan, Zeynep ; Babaoglu, Ozalp ; Kodama, Yuetsu
Pubblicato in: 2024
Editore: Universtià di Bologna

Assessing Tenstorrent’s RISC-V MatMul Acceleration Capabilities

Autori: Hiari Pizzini Cavagna, Daniele Cesarini, Andrea Bartolini
Pubblicato in: 2025
Editore: Cornell University