Periodic Reporting for period 1 - Rising STARS (RISE International Network for Solutions Technologies and Applications of Real-time Systems)
Período documentado: 2020-02-01 hasta 2023-11-30
Cyber-Physical Systems (CPS), traditionally composed of embedded computing platforms and sensors/actuators in charge of coordinating and integrating cyber (computational) and physical processes, are increasingly required to exhibit a high degree of autonomy and intelligence to address societal, industrial and scientific challenges across multiple application domains for autonomous and safe mobility, sustainable production and smart manufacturing, giant scientific installations, etc .. This trend is however challenging the development of the newest CPS, which increasingly require high-performance capabilities to coordinate and integrate very complex computational and physical processes, while requiring to fulfil non-functional constraints such as energy efficiency or real-time response imposed due to interactions between the cyber and physical world. One prominent example of the new CPS are the next generation of giant scientific equipment such as the European Extremely Large Telescope (ELT) or the Square Kilometer Array (SKA). High Performance Computing (HPC) has become a critical component to optimize the design of these science mega-factories and produce high-level scientific data for hundreds of research teams.
The main goal of Rising STARS is to extend parallel programming frameworks for the development and execution of advanced large-scale CPS with HPC and real-time requirements. In order to meet the real-time constraints while assimilating vast amounts of data, the mechanism of acquiring these data have to be interleaved with the computations in an optimized way while keeping three main characteristics of these CPS: flexible, deterministic and high performance.
To that end, we have defined 6 core objectives:
* Objective #1: enable a versatile and efficient data acquisition platform based on FPGA.
* Objective #2: expose data acquisition mechanisms in the programming model.
* Objective #3: introduce real-time oriented features in the programming model.
* Objective #4: develop a technology demonstration platform for Adaptive Optics (AO) on giant telescopes.
* Objective #5: lead a case study for data processing on SKA.
* Objective #6: leverage similar technological solutions for other applications.
Our strategy is based on a strong interaction between academic and industrial partners. Beyond technological developments, we will rely on an in situ implementation of our solutions on challenging prototypes for large scale scientific experiments with world wide visibility.
The main goal of Rising STARS is to extend parallel programming frameworks for the development and execution of advanced large-scale CPS with HPC and real-time requirements. In order to meet the real-time constraints while assimilating vast amounts of data, the mechanism of acquiring these data have to be interleaved with the computations in an optimized way while keeping three main characteristics of these CPS: flexible, deterministic and high performance.
To that end, we have defined 6 core objectives:
* Objective #1: enable a versatile and efficient data acquisition platform based on FPGA.
* Objective #2: expose data acquisition mechanisms in the programming model.
* Objective #3: introduce real-time oriented features in the programming model.
* Objective #4: develop a technology demonstration platform for Adaptive Optics (AO) on giant telescopes.
* Objective #5: lead a case study for data processing on SKA.
* Objective #6: leverage similar technological solutions for other applications.
Our strategy is based on a strong interaction between academic and industrial partners. Beyond technological developments, we will rely on an in situ implementation of our solutions on challenging prototypes for large scale scientific experiments with world wide visibility.
The project was suspended during almost 2 years during the Covid-19 pandemic and restarted in January 2022 after an amendment to the initial grant agreement was signed. Objectives have been mostly achieved: several papers outlining preliminary results from the collaboration have been published and two training events were co-sponsored by the project, including a hackathon. The first two milestones were achieved.
Regarding objectivs #1: in collaboration with NVIDIA and Thales, a preliminary technology evaluation was performed under the form of a high performance data acquisition prototype implemented on a SKA pathfinder in France (NenuFAR)
Regarding objective 2, new Parallel Programming Models (PPM) extensions to support explicit data transfers and acquisitions have been proposed and validated. In particular, a new extension to OpenMP to support event-driven model including new clauses event, periodic and sporadic associated to the task directive.
Regarding objective #3, new extensions to OpenMP have been developed to better characterize the parallel structure of a regions with the two-fold objective of: reducing/eliminating the task overhead on CPU/GPU and including the information needed for real-time guarantees
Regarding objective #4, the project focused on an innovative development: use Deep Neural Networks for AO control with PhD project between BSC and Observatoire de Paris. This collaboration led to several publications as well as the implementation of a co-tutelle agreement.
Regarding objective #5, proposed work on Data Processing Units (DPU) was implemented as an alternative to the FPGA data interface and prototyping results were published, including the development & integration of Pulsar / Fast Radio Burst real-time detection pipeline.
Regarding objective #6, after a first phase of specification capture for Critical Real-Time Embedded System (CRTES), a reference implementation for a typical CRTES (Adaptive beam forming) was developed as well as a case study for porting the AO application to embedded platforms (Real-time on embedded NVIDIA GPU) that led to a publication.
The project has led to a number of publications (5 so far as reported on the participant portal).
Regarding objectivs #1: in collaboration with NVIDIA and Thales, a preliminary technology evaluation was performed under the form of a high performance data acquisition prototype implemented on a SKA pathfinder in France (NenuFAR)
Regarding objective 2, new Parallel Programming Models (PPM) extensions to support explicit data transfers and acquisitions have been proposed and validated. In particular, a new extension to OpenMP to support event-driven model including new clauses event, periodic and sporadic associated to the task directive.
Regarding objective #3, new extensions to OpenMP have been developed to better characterize the parallel structure of a regions with the two-fold objective of: reducing/eliminating the task overhead on CPU/GPU and including the information needed for real-time guarantees
Regarding objective #4, the project focused on an innovative development: use Deep Neural Networks for AO control with PhD project between BSC and Observatoire de Paris. This collaboration led to several publications as well as the implementation of a co-tutelle agreement.
Regarding objective #5, proposed work on Data Processing Units (DPU) was implemented as an alternative to the FPGA data interface and prototyping results were published, including the development & integration of Pulsar / Fast Radio Burst real-time detection pipeline.
Regarding objective #6, after a first phase of specification capture for Critical Real-Time Embedded System (CRTES), a reference implementation for a typical CRTES (Adaptive beam forming) was developed as well as a case study for porting the AO application to embedded platforms (Real-time on embedded NVIDIA GPU) that led to a publication.
The project has led to a number of publications (5 so far as reported on the participant portal).
Because of the multidisciplinary nature of the Rising STARS research and development program, the outcomes of the project are expected to impact several domains from HPC specific topics to critical real-time systems. Moreover, thanks to a strong partnership between academic and industrial partners, the various case studies will help to strengthen the competitiveness and growth of European companies by providing competitive technology demonstration, with a world-wide visibility, promoting the potential of these innovations to enable world-leading science. This co-design approach will eventually lead to the delivery of these innovations to the market with the added value of a credible path to deliver an application specific solution using them with access to the corresponding reusable tools / extensions.
On the academic side, the Rising STARS partners, through the development of new solutions to drive AO systems, at the core of the telescope operations, is thus expected to have a critical impact on the astronomical instrumentation community. On top of this, the case study on SKA and the strong collaboration with Australian partners already involved in this project would provide European radio-astronomer a prime access to this major facility and would provide European high-tech companies, such as Kalray and Thales, new business opportunities and prime exposure on a world leading scientific facility.
On the private sector side, we expect significant outcomes for the various partners of the project. Concerning Microgate, the marketing strategy foresees to address first the market niche related to AO and approach other markets, in particular the fast image processing required by biomedical science and by some industrial and automotive applications. Many Thales Group’s operational Business Units (GBUs) need to achieve HPC and critical real-time all at once on embedded systems. The convergence of these needs, that should be addressed by the evolution of a standardised, portable and friendly-user parallel programming model could be quickly and efficiently exploited by GBUs. Rising STARS represents also an excellent opportunity to allow KALRAY to increase the real-time capabilities of the MPPA by enhancing its software development kit (based on OpenMP) with new dedicated real-time and data transfer and acquisition extensions. Finally, the Rising STARS project will contribute significantly to the business growth of ArianeGroup in Australia and to and to identify new business opportunities with major satellites operators in the area thanks to enhanced sky coverage enabled by new cooperations with Australian scientists and Universities.
On the academic side, the Rising STARS partners, through the development of new solutions to drive AO systems, at the core of the telescope operations, is thus expected to have a critical impact on the astronomical instrumentation community. On top of this, the case study on SKA and the strong collaboration with Australian partners already involved in this project would provide European radio-astronomer a prime access to this major facility and would provide European high-tech companies, such as Kalray and Thales, new business opportunities and prime exposure on a world leading scientific facility.
On the private sector side, we expect significant outcomes for the various partners of the project. Concerning Microgate, the marketing strategy foresees to address first the market niche related to AO and approach other markets, in particular the fast image processing required by biomedical science and by some industrial and automotive applications. Many Thales Group’s operational Business Units (GBUs) need to achieve HPC and critical real-time all at once on embedded systems. The convergence of these needs, that should be addressed by the evolution of a standardised, portable and friendly-user parallel programming model could be quickly and efficiently exploited by GBUs. Rising STARS represents also an excellent opportunity to allow KALRAY to increase the real-time capabilities of the MPPA by enhancing its software development kit (based on OpenMP) with new dedicated real-time and data transfer and acquisition extensions. Finally, the Rising STARS project will contribute significantly to the business growth of ArianeGroup in Australia and to and to identify new business opportunities with major satellites operators in the area thanks to enhanced sky coverage enabled by new cooperations with Australian scientists and Universities.