Periodic Reporting for period 2 - RED-SEA (Network Solution for Exascale Architectures)
Periodo di rendicontazione: 2022-10-01 al 2024-03-31
The interconnection networks serve as the backbone of Exascale systems and play a crucial role in the overall performance of Exascale systems. To meet the demands of Exascale computing, these networks must support high node counts, massively parallel processing systems, efficient connection to the data-centre network, and emerging data-centric and AI-related applications. Additionally, they need to incorporate features such as efficient network resource management, in-network computing, and power-efficient support for accelerators and compute units.
RED-SEA was built upon the European interconnect BXI (BullSequana eXascale Interconnect), together with standard and mature technology (Ethernet) and previous EU-funded initiatives to provide a competitive and efficient network solution for the exascale era and beyond. This involved developing the key IPs and the software environment that delivered:
- scalability, while maintaining an acceptable total cost of ownership and power efficiency;
- virtualization and security, to allow various applications to efficiently and safely share an HPC system;
- Quality-of-service and congestion management to make it possible to share the platform among users and applications with different demands;
- reliability at scale, because fault tolerance is a key concern in a system with a very large number of components;
- support of high-bandwidth low-latency HPC Ethernet, as HPC systems increasingly need to interact securely with the outside world, including public clouds, edge servers or third party HPC systems;
- support of heterogeneous programming model and runtimes to facilitate the convergence of HPC and HPDA;
- support for low-power processors and accelerators.
At the conclusion of the project, we have fully achieved our objectives. After three years of dedicated effort, we have successfully advanced the state of interconnect network technologies.
New advances in interconnect technologies can help establish Europe in retaining its position as a leader pioneer in exascale supercomputing. This will bring benefits across a range of fields, from climate modelling to drug discovery.
RED-SEA was built upon the European interconnect BXI (BullSequana eXascale Interconnect), together with standard and mature technology (Ethernet) and previous EU-funded initiatives to provide a competitive and efficient network solution for the exascale era and beyond. This involved developing the key IPs and the software environment that delivered:
- scalability, while maintaining an acceptable total cost of ownership and power efficiency;
- virtualization and security, to allow various applications to efficiently and safely share an HPC system;
- Quality-of-service and congestion management to make it possible to share the platform among users and applications with different demands;
- reliability at scale, because fault tolerance is a key concern in a system with a very large number of components;
- support of high-bandwidth low-latency HPC Ethernet, as HPC systems increasingly need to interact securely with the outside world, including public clouds, edge servers or third party HPC systems;
- support of heterogeneous programming model and runtimes to facilitate the convergence of HPC and HPDA;
- support for low-power processors and accelerators.
At the conclusion of the project, we have fully achieved our objectives. After three years of dedicated effort, we have successfully advanced the state of interconnect network technologies.
New advances in interconnect technologies can help establish Europe in retaining its position as a leader pioneer in exascale supercomputing. This will bring benefits across a range of fields, from climate modelling to drug discovery.
The main outcomes of the project are summarised as following:
• One significant outcome lies in the advancement of the European Interconnect BXI, with a focus on enhancing the current version (BXIv2) and laying the groundwork for its next generation (BXIv3) through thorough design and specification work.
• Another achievement of the project is its exploration of new, efficient network resource management schemes. We have enhanced collective operations, congestion control, and adaptive routing.
• We have extended the European Interconnection networks ecosystem through the expansion of the usage of BXI networks, adding new IPs blocs in the Network Interfaces dealt with in the RED-SEA project and the adaptation of applications to the Network Interfaces.
• In addition, we have undertaken several additional developments to enhance European technology Intellectual Properties (IPs) within the field of interconnection networks.
• Finally, we have disseminated the outcomes through academic publications, open access resources and a number of events.
The diversity of partners is the guarantee of a great diversity of exploitations: products, patents, IP portfolio, open-source tools. As results, we have identified and developped 21 RED-SEA Exploitable Results and 2 patents from all consortium partners. Over 50% (Eleven) of these results are product-related, with some already integrated into commercial products, the others will be implemented in a commercial product within two years.
For dissemination, the collaboration with the other SEA projects and with other EuroHPC projects was essential to give visibility to RED-SEA. This collaboration initiated very early in the project made it possible for RED-SEA to have a significant and visible presence in many events (more than we initially anticipated), which would not have been possible on our own. Concerning dissemination, the very limited number of public deliverables hampered the dissemination of the project's findings - due to the sheer lack of material, especially as scientific publications are often derived from deliverables. This high level of confidentiality obviously has to do with the fact that the project is largely based on a commercial product with proprietary IP – an advantage for exploitation, but a problem for dissemination. However, RED-SEA managed to publish 19 academic papers, all available in Open Access. Publications are listed in the About section of the RED-SEA website: https://redsea-project.eu/about/#publications .
• One significant outcome lies in the advancement of the European Interconnect BXI, with a focus on enhancing the current version (BXIv2) and laying the groundwork for its next generation (BXIv3) through thorough design and specification work.
• Another achievement of the project is its exploration of new, efficient network resource management schemes. We have enhanced collective operations, congestion control, and adaptive routing.
• We have extended the European Interconnection networks ecosystem through the expansion of the usage of BXI networks, adding new IPs blocs in the Network Interfaces dealt with in the RED-SEA project and the adaptation of applications to the Network Interfaces.
• In addition, we have undertaken several additional developments to enhance European technology Intellectual Properties (IPs) within the field of interconnection networks.
• Finally, we have disseminated the outcomes through academic publications, open access resources and a number of events.
The diversity of partners is the guarantee of a great diversity of exploitations: products, patents, IP portfolio, open-source tools. As results, we have identified and developped 21 RED-SEA Exploitable Results and 2 patents from all consortium partners. Over 50% (Eleven) of these results are product-related, with some already integrated into commercial products, the others will be implemented in a commercial product within two years.
For dissemination, the collaboration with the other SEA projects and with other EuroHPC projects was essential to give visibility to RED-SEA. This collaboration initiated very early in the project made it possible for RED-SEA to have a significant and visible presence in many events (more than we initially anticipated), which would not have been possible on our own. Concerning dissemination, the very limited number of public deliverables hampered the dissemination of the project's findings - due to the sheer lack of material, especially as scientific publications are often derived from deliverables. This high level of confidentiality obviously has to do with the fact that the project is largely based on a commercial product with proprietary IP – an advantage for exploitation, but a problem for dissemination. However, RED-SEA managed to publish 19 academic papers, all available in Open Access. Publications are listed in the About section of the RED-SEA website: https://redsea-project.eu/about/#publications .
The industry/academia collaboration was a strong point for the success of the project, with a large-industry partner (BULL), 4 SMEs (EXTOLL, Exapsys, Exact-lab and ParTec) and prominent research and academia partners (FORTH, INFN, CEA, JSC, EthZ, UPV, UCLM). This was instrumental in ensuring that the homegrown technologies developed within the project are or will be actually rolled out in European and international systems and further projects:
• all RED-SEA partners have declared Exploitable Results,
• all “commercial” RED-SEA partners have already integrated or will integrate some outcomes of the project in their products within two years,
• all academic RED-SEA partners are already using or planning to use some outcomes of the project for further research projects.
The top RED-SEA impact is without doubt that the project provided the foundation to develop and implement the next generation BXI, refine and optimise performance based on feedback, testing, and real-world use cases to ensure its effectiveness and efficiency. BXIv3 is planned to be integrated in the EUPEX project, which is a Pilot system for Exascale systems with European technologies - EUPEX is a project funded by EuroHPC JU. Mastering Interconnect technologies is an advantage for Europe in terms of sovereignty and economic return. The IP blocs developed in the RED-SEA project have significantly enriched the European technology portfolio in the field of interconnection networks. These European IP blocs are poised to serve as strong candidates for integration into Exascale systems and future computing architectures.
Beyond the IP portfolio for European Interconnect network technologies, it was equally important to reinforce the skills necessary for interconnect design. The consortium therefore made important efforts to share and perpetuate the knowledge, methodology and tools used/developed by the project for interconnection network simulation and HW prototyping, e.g. the tools that allowed partners to collect network traces and analyse the network traffic prior to the development of the mechanisms for collective operations, congestion control, and adaptive routing. The Consortium members shared this knowledge through many workshops and seminars.
We hope that the BXI roadmap will be a long-term plan thanks to the outcomes of the RED-SEA project and future projects. The long-term legacy of the interconnect network technologies project is expected to result in transformative societal benefits, including scientific advancement, improved healthcare, environmental sustainability, energy efficiency, national security, and education. By harnessing the power of Exascale computing, we aspire to create a more prosperous, equitable, and sustainable future for all.
• all RED-SEA partners have declared Exploitable Results,
• all “commercial” RED-SEA partners have already integrated or will integrate some outcomes of the project in their products within two years,
• all academic RED-SEA partners are already using or planning to use some outcomes of the project for further research projects.
The top RED-SEA impact is without doubt that the project provided the foundation to develop and implement the next generation BXI, refine and optimise performance based on feedback, testing, and real-world use cases to ensure its effectiveness and efficiency. BXIv3 is planned to be integrated in the EUPEX project, which is a Pilot system for Exascale systems with European technologies - EUPEX is a project funded by EuroHPC JU. Mastering Interconnect technologies is an advantage for Europe in terms of sovereignty and economic return. The IP blocs developed in the RED-SEA project have significantly enriched the European technology portfolio in the field of interconnection networks. These European IP blocs are poised to serve as strong candidates for integration into Exascale systems and future computing architectures.
Beyond the IP portfolio for European Interconnect network technologies, it was equally important to reinforce the skills necessary for interconnect design. The consortium therefore made important efforts to share and perpetuate the knowledge, methodology and tools used/developed by the project for interconnection network simulation and HW prototyping, e.g. the tools that allowed partners to collect network traces and analyse the network traffic prior to the development of the mechanisms for collective operations, congestion control, and adaptive routing. The Consortium members shared this knowledge through many workshops and seminars.
We hope that the BXI roadmap will be a long-term plan thanks to the outcomes of the RED-SEA project and future projects. The long-term legacy of the interconnect network technologies project is expected to result in transformative societal benefits, including scientific advancement, improved healthcare, environmental sustainability, energy efficiency, national security, and education. By harnessing the power of Exascale computing, we aspire to create a more prosperous, equitable, and sustainable future for all.