Predictable Parallel Computing

Rising STARS

Title: RISE International Network for Solutions Technologies and Applications of Real-time Systems

Programme: MSCA-RISE - Marie Skłodowska-Curie Research and Innovation Staff Exchange (RISE)

Start date: 2020-02-01 End date: 2024-01-01

Description and partners

The main goal of Rising STARS is to enable a parallel programming framework for the development and execution of advanced large-scale Cyber Physical Systems (CPS) with High Performance Computing (HPC) and real-time requirements. Overall, there is an urgent necessity to develop run-time parallel frameworks, compatible with HPC, capable of guaranteeing that decisions made at run-time maintains the guarantees about system correctness and timing behavior. These new run-time capabilities however, cannot preclude the ability of run-times to dynamically adapt the execution to new working conditions or changing modes of operation of CPS to maximize the utilization and performance capabilities of parallel heterogeneous architectures.

A key element of the Rising STARS framework will be the incorporation of a unified, efficient and highly configurable data acquisition strategy fully integrated in the parallel programming models with the objective of improving productivity in CPS software development. Exposing the data-acquisition to the programmer (by including it into the parallel programming model) is also key to overlap data-transfers with computation. Another objective of the project is to add this capability in existing programming models for HPC and to investigate new parallel programming extensions to allow developers to define the real-time properties of the system in terms of periodicity and timing constraints. Finally, one of our main objectives is to implement several demonstration platforms to promote the main technological developments of this R&I action and their performance under realistic conditions, including Adaptive Optics for giant telescopes and SSA experiments, data processing for SKA, and critical real-time embedded systems.

Cordinator: Observatoire de Paris

Participant: Barcelona Supercomputing Center, Thales SA, Microgate SRL, ArianeGroup SAS, Kalray SA, NVIDIA Switzerland AG, The Australian National University, Swinburne University Technology, Commonwealth Scientific and Industrial Research Organization, Inter- university Research Institute Corporation National Institutes of Natural Sciences

AMPERE

Title: A Model-driven Programming framework supporting highly parallel and EneRgy-Efficient execution

Programme: RIA - Research and Innovation action

Start date: 2020-01-01 End date: 2022-12-01

Description and partners

Complex, dependable and physically-entangled systems of systems must be supported by innovations to allow a significant reduction of the cost and complexity of system design targeting computing platforms composed of parallel heterogeneous architectures. Software development is one key challenge, as current programming tools do not fully support emerging processor architectures. Parallel and heterogeneous platforms are difficult to program and even more to optimize for the multiple conflicting criteria imposed by applications, such as performance, energy efficiency, real-time response, resiliency and fault tolerance.

AMPERE addresses this challenge by incorporating model-driven engineering (MDE) as the key element for the construction of complex software architectures. MDE enables to efficiently capture system's functional and non-functional requirements, including multiple conflicting requirements, as well as enabling the use of domain specific model-driven languages (DSML) to further refine the description of cyber/physical interactions.

The vision of AMPERE is that there is a clear necessity of developing a new generation of code synthesis methods and tools capable to implement correct-by-construction systems, in which the constraints captured by the system model are efficiently transformed to the parallel programming models supported by the underlying parallel heterogeneous platform, whilst providing the level of performance required. Moreover, AMPERE will provide computing software composed of a set of advanced run-time methods implementing monitoring and dynamic reconfiguration techniques, that will support the parallel execution to improve the overall system's efficiency, and guarantee that the non-functional requirements capture by the DSML are fulfilled. AMPERE advances will be integrated in a set of ready-to-use tools and libraries, and validated through demonstration in two reference applications, from automotive and railway domains.

Cordinator: Barcelona Supercomputing Center

Participant: Instituto Superior de Engenharia do Porto, Scuola Superiore di Studi Universitari e di Perfezionamento Sant'Anna, Evidence Srl, Bosch, Thales SA, Thales Italy, and SYSGO

DEEPHEALTH

Title: Deep-Learning and HPC to Boost Biomedical Applications for Health

Programme: IA - Innovation action

Start date: January 2019 End date: January 2022

Website: https://deephealth-project.eu/

Description and partners

The aim of DeepHealth is to offer a unified framework completely adapted to exploit underlaying heterogeneous HPC and Big Data architectures; and assembled with state-of-the-art techniques in Deep Learning and Computer Vision. In particular, DeepHealth framework is envisioned to tackle real needs of the health sector and facilitate the daily work of medical personnel and the expert users in terms of image processing and the use and training of predictive models without the need of combining numerous tools. To this end, the project will combine High-Performance Computing (HPC) infrastructures with Deep Learning (DL) and Artificial Intelligence (AI) techniques to support biomedical applications that require the analysis of large and complex biomedical datasets and thus, new and more efficient ways of diagnosis, monitoring and treatment of diseases. Moreover, two new libraries, the European Distributed Deep Learning Library (EDDLL) and the European Computer Vision Library (ECVL), will be developed and incorporated in the DeepHealth framework for manipulating and processing the images in a more efficient way and thus, for increasing the productivity of professionals working on biomedical images. The resulting enhanced diagnosis will significantly improve the health service provided to the society, making public health systems more efficient and profitable for everyone.

Cordinator: Everis Spain SL (EVR)

Participant: Universitat Politecnica de Valencia (UPV), Philips Medical Systems Nederland BV (PHILIPS), Siveco Romania SA (SIVECO), Wings ICT Solutions Information & Communication Technologies IKE (WINGS),Thales Six gts France sas (THALES SIX), Commissariat A L Energie Atomique et aux Energies Alternatives (CEA), Barcelona Supercomputing Center - Centro Nacional de Supercomputacion (BSC), Pro Design Electronic GmbH (PRO DESIGN) Karolinska Institutet (KI), Fundacion para el Fomento de la Investigacion Sanitaria y Biomedica de la Comunitat Valenciana (FISABIO), Azienda Ospedaliera Citta della salute e Della Scienza di Torino (CDSS), Ecole Polytechnique Federale de Lausanne (EPFL), Centre Hospitalier Universitaire Vaudois (CHUV), Tree Technology SA (TREE), Otto-von-Guericke-Universitaet Magdeburg (OVGU), Stelar Security Technology Law Research UG (STELAR), Spitalul Clinic Prof DR Theodor Burghele (SCHTB), Universita degli studi di Torino (UNITO), Universita degli studi di Modena e Reggio Emilia (UNIMORE), Centro di Ricerca, Sviluppo e studi superiori in Sardegna Società a Responsabilità Limitata (CRS4)

ELASTIC

Title: A Software Architecture for Extreme-ScaLe Big-Data AnalyticS in Fog CompuTIng ECosystems

Programme: RIA - Research and Innovation action

Start date: 2018-12-01 End date: 2021-11-01

Website: https://elastic-project.eu/

Description and partners

ELASTIC will develop a software architecture incorporating a new elasticity concept, that will enable smart systems to satisfy the performance requirements of extreme-scale analytics workloads. The new elasticity concept will efficiently distribute the workloads across the compute continuum, whilst guaranteeing real-time, energy, communication quality and security non-function properties inherited from the system domain.

The vision of ELASTIC is that by extending the elasticity concept across the compute continuum in a fog computing environment, combined with the usage of advanced hardware architectures at the edge side, can significantly increase the capabilities of the extreme-scale analytics integrating both responsive data-in-motion and latent data-at-rest analytics into a single solution.

ELASTIC will consider a realistic yet visionary smart mobility use-case, which will elaborate of huge amounts of data coming from a large set of IoT sensors distributed along the Florence tramway network. ELASTIC will adopt a very innovative federated/distributed fog architecture, supporting elasticity across the compute continuum whilst fulfilling real-time, energy, communication and secure properties.

Cordinator: Barcelona Supercomputing Center

Participant: Ikerlan, Instituto Superior De Engenharia Do Porto, Information Catalyst for Enterprise Ltd, SIXSQ SARL, THALES SA, THALES Italia SPA, Gestione Ed Eserizio del Sistema Tranviario SPA, Citta Metropolitana di Firenze

Hp4s

Title: High Performance Parallel Payload Processing for Space

Start date: 2018-07-01 End date: 2018-06-01

Description and partners

The payload data processing applications have increasing processing needs. Parallel processing with multi-core General Purpose Processors (GPP) is an interesting solution for meeting these needs, and many-core architectures are very promising. The HP4S study aims at evaluating the viability of such architectures and the parallel programming framework OpenMP for payload data processing applications.

Cordinator: Airbus Defence and Space

Participant: Barcelona Supercomputing Center

CLASS

Title: Edge and Cloud Computation: A Highly Distributed Software for Big Data Analytics

Programme: RIA - Research and Innovation action

Start date: 2018-01-01 End date: 2020-12-01

Website: https://class-project.eu/

Description and partners

CLASS aims to develop a novel software architecture framework to help big data developers to efficiently distributing data analytics workloads along the compute continuum (from edge to cloud) in a complete and transparent way, while providing sound real-time guarantees. This ability opens the door to the use of big data into critical real-time systems, providing to them superior data analytics capabilities to implement more intelligent and autonomous control applications.

The capabilities of the CLASS framework will be demonstrated on a real smart-city use case in the City of Modena, featuring a heavy sensor infrastructure to collect real-time data across a wide urban area, and three connected vehicles equipped with heterogeneous sensors/actuators and V2X connectivity to enhance the driving experience.

Cordinator: Barcelona Supercomputing Center

Participant: Atos, Comune di Modena, IBM Research - Haifa, Università degli studi di Modena e Reggio Emilia (UNIMORE) and Maserati

Finished projects

Increasing the Guaranteed Performance in Many-core Heterogeneous Architectures

From: 2016 to 2017

Description and partners

Many-core heterogeneous architectures are rapidly evolving, providing to embedded systems a unprecedented level of performance required to cope with current and future system challenges. However, in order to efficiently exploit the massively parallel computation capabilities of these architectures, it is mandatory to tame the complexity of parallel programming.

This project investigates parallel programming models, scheduling and timing estimation techniques to obtain a high-performance and tight response-time bounds of parallel computation. The proposed approach is to be demonstrated on automotive Advances Driver Assistance Systems (ADAS).

Cordinator: Barcelona Supercomputing Center.

Participant: DENSO AUTOMOTIVE Deutschland GmbH.

Parallel Programming Models for Space Systems

From: 2015 to 2016

Description and partners

This projects aims to demonstrate the benefits of using OpenMP v4.0 tasking model in space systems in order to improve performance speed-up and increase programmability, while still providing timing analyzability. To do so, this project will identify the challenges that future implementations of OpenMP4 must address so it can be applied to space systems.

This project is an ESA contract from the Innovation Triangle Initiative framework (proof-of-concept type).

Cordinator: Barcelona Supercomputing Center.

Participant: Evidence Srl.

P-SOCRATES

Title: Parallel Software Framework for Time-Critical many-core Systems

Programme: CP - Collaborative project (generic)

From: 2013 to 2016

Website: https://p-socrates.github.io/

Description and partners

P-SOCRATES proposes an integrated framework for executing workload-intensive parallel applications with real-time requirements on top of next-generation commercial-off-the-shelf (COTS) platforms based on many-core accelerated architectures. The project will investigate new HPC techniques that fulfil real-time requirements. The main sources of indeterminism will be identified, proposing efficient mapping and scheduling algorithms, along with the associated timing and schedulability analysis, to guarantee the real-time and performance requirements of the applications.

Cordinator: Instituto Superior de Engenharia do Porto

Participant: Barcelona Supercomputing Center, University of Modena, ETH Zurich, Evidence Srl, Active Technologies Srl, ATOS Spain SL.