25 Sep 2015 Brussels - The European Commission published the draft of the Horizon 2020 Work Programme 2016-2017. Although not final, real changes are very unlikely. Horizon 2020 is the European Research and Development programme. It supprts R&D on a European scale, supplementing national R&D programmes. Every two years, a detailed work programme is published, describing in detail which amount of grant money will be available for which type of R&D. HPC is one of the topics that are in focus. There is a wide range of topics in different sub programmes of the two year 2016-2017 Work Programme. We collect them to give a comprehensive overview. Please keep in mind this is a two-year plan, so it might well be that the focus shifts in 2018-2019. The target is not yet exascale in the next two years, but 250 Petaflop/s at 15MwW should be achievable in 2019. Meanwhile both the PRACE supercomputing infrastructure will be supported and The Human Brain project will receive 25 million to invest in a specialized 50 Petaflop/s supercomputer.
The main HPC topics in focus in the next two years are:
There are some other areas that the Commission funds which are strongly related. This includes distributed computing, Cloud computing (including the European Science Cloud Initiative) and Big Data innovation. We will not cover these in this article.
The PRACE supercomputing infrastructure is one of the infrastrucutures for "advanced communities", in particular the Mathematics and ICT community. Each of these advanced communities can get up to 10 million euro for integrating activities that include:
1. Networking activities, to foster a culture of co-operation between research infrastructures, scientific communities, industries and other stakeholders as appropriate, and to help develop a more efficient and attractive European Research Area;
2. Trans-national access or virtual access activities, to support scientific communities in their access to the identified key research infrastructures;
3. Joint research activities, to improve, in quality and/or quantity, the integrated services provided at a European level by the infrastructures.
The focus for the supercomputer activity is on providing these integrating activities for facilitating access to HPC (High Performance Computing) centres. This activity aims at furthering the services harmonisation and enhancement of national and regional High Performance Computing Centres of pan-European interest and at enlarging the European HPC user base preparing it to the use of the top end HPC resources such as PRACE (Partnership for Advanced Computing in Europe). It will widen trans-national access to HPC resources across different disciplines and for a wide range of applications including advanced simulation and modelling.
In addition there is support envisagned by the European Commission for the support to the next implementation phase of pan-European High Performance Computing infrastructure and services (PRACE). So this focuses more on the supercomputer systems themselves.
The Commission expects that in order to create a world-class pan-European High Performance Computing (HPC) infrastructure and to provide state-of-the-art services with access by users, independently of their location, the HPC resources need to be further pooled, integrated and rationalised. Horizon 2020 is a key element to support the European strategy on High Performance Computing which sets the way forward, in particular regarding the access to the best supercomputing facilities and services for both industry and academia.
What is the expected outcome? The funded project should:
1. Provide a seamless and efficient Europe-wide Tier-0 service to users, based on
promoting research excellence and innovation. Supercomputer time will be allocated on the basis of a peer-review system. There must be specific services adapted to the needs of users, including ESFRI projects, Horizon 2020 projects/programmes, large institutional users or industry.
2. Carry out activities (training, service prototyping, software development etc.) that build on national HPC capabilities (Tier-1) and are necessary to support Tier-0 services or a functional European HPC ecosystem.
3. Ensure openness to new user communities and new applications and promote industrial take-up of HPC services in particular by SMEs.
4. Implement inclusive and equitable governance and a flexible business model to ensure long term financial sustainability; the business model should allow financial or in-kind contributions by research projects/programmes, institutions, industry and regions or countries; based on an auditable cost model for the operation of HPC Centres providing European services with different financing sources;
5. Develop and maintain the strategy for the deployment of a rich HPC environment of world-class systems with different machine architectures evolving towards exascale.
6. Working in synergy with the European Technology Platform for HPC (ETP4HPC) and the Centres of Excellence in HPC applications. In particular, the mechanisms to be put in place by PRACE for the provision of technical specifications to guide research activities for future exascale prototypes and systems and the transfer of know-how for the use of HPC.
7. Design and execute training and skills development programmes.
8. Develop an international cooperation policy and associated activities in the area of HPC aiming at systems interoperation.
The Commission expects this initiative will help Europe staying at the forefront of scientific breakthroughs and innovation. It will increase the number of industrial organisations (in particular SMEs), EU projects and institutional users benefiting from access to services including training in HPC. Benefits will also translate in increased investment in HPC infrastructure in Europe.
The Future and Emerging Technologies subprogramme is intended to kick start new technologies in Europe. This can range from novel ideas to the large flagship projects on Graphene and the Human Brain.
A complete chapter in Future and Emerging Technologies is devoted to HPC: "FET Proactive - High Performance Computing". In two years' time the EC expects to fund this area with 85 million euro. The European Commission considers HPC as an important technology for Europe.
The focus is on advanced HPC towards exascale. There are three main topics:
Co-design of HPC systems and applications
Co-design of HPC systems is where the Commission expects some big projects, somewhere between 10 and 20 million euro each.
HPC platforms should be developed through a strong co-design approach driven by large applications and in close cooperation with the scientific disciplines and stakeholders concerned. The designs of the systems developed in these projects must respond to critical demands of energy efficiency, scale, resilience, programmability and support for various classes of applications including extreme-data applications.
Funded projects are expected to have innovative and ground-breaking approaches to system architectures targeting extreme scale, power-efficient and highly resilient platforms with emphasis on balanced compute and data access characteristics.
It is not only about exascale processing. Special attention should be given by the projects to extreme data processing requirements. As the title suggests they must have a strong co-design approach.
Energy efficiency and reduced energy-to-solution for the selected applications are also important. Possible strategies for improving energy efficiency suggested by the European Commission include: reducing PUE (Power Usage Effectiveness), designing of cost-efficient approaches to the reuse of thermal energy, reducing the amount of energy spent for communication and data movement. Fault tolerance is also of great importance.
Projects should provide analytical or simulation models that allow to extrapolate the sustained performance on the given architecture for HPC systems. The target system architectures must scale to at least 100 Petaflop/s and, for compute-centric workloads, a target of 15MW for 250 Petaflop/s peak performance in 2019 is possible, the Commission suggests.
Because it is a co-design of a system and its applications, project partners should explain how the scalability and energy-efficiency targets are achieved for the considered co-design applications. APIs and interfaces between applications and underlying middleware, run-time and operating systems, i.e. all application-aspects impacting the underlying system design are included in this topic.
But it is not all about a pencil and paper exercise. Projects should demonstrate their achievements in integrated pre-exascale prototypes.
Transition to Exascale Computing
Projects in "Transition to Exascale computing" are expected to be smaller and more focused than the ones in "Co-design of HPC systems and applications", typically between 2 and 4 million euro.
What the Commission wants projects in this area to address is to take advantage of the full capabilities of exascale computing. This should be done in particular through high-productivity programming environments, system software and management, exascale I/O and storage in the presence of multiple tiers of data storage, supercomputing for extreme data and emerging HPC use modes, mathematics and algorithms for extreme scale HPC systems for existing or visionary applications. Applications could include data-intensive and extreme data applications in scientific areas such as physics, chemistry, biology, life sciences, materials, climate, and geosciences.
There are 5 topics on which projects should focus:
1. High productivity programming environments for exascale
The main target is to simplify application software development for large- and extreme-scale systems. Key aspects include managing data transfers, data locality and memory management, including support for heterogeneous and reconfigurable systems as well as dealing with inter-application dynamic load balancing, and adapting to changes in the number of processors. Also support for domain-specific languages (DSLs) frameworks should be considered.
2. Exascale system software and management
The main target here is to advance the state of the art in system software and management for node architectures that will be drastically more complex. These architectures' resource topology and heterogeneity will require OS and runtime enhancement, such as data aware scheduling.
In the area of hardware abstraction, the goal is to address run time handling of all types of resources (cores, bandwidth, logical and physical memory or storage) and controls for optimised data coherency, consistency and data flow. For applications, the idea is to address new multi-criteria resource allocation capabilities and interaction during task execution, with the aim to improve resilience, interactivity, power and efficiency.
3. Exascale I/O and storage in the presence of multiple tiers of data storage
The goal of this topic is to address exascale I/O systems expected to have multiple tiers of data storage technologies, including non-volatile memory.
4. Supercomputing for Extreme Data and emerging HPC use modes
This topic is about HPC architectures for real-time and in-situ data analytics that are required to support the processing of large-scale and high velocity real-time data. So the Commission is looking for projects that include support for real-time in-memory analysis of different data structures, direct processing of compressed data and appropriate benchmarking method for performance analysis.
Apart from interactive 3-D visualisation of large-scale data, interactive supercomputing is considered as well. This is required to execute complex workflows for urgent decision making for instance in the field of critical clinical diagnostics or spread of diseases.
5. Mathematics and algorithms for extreme scale HPC systems and applications working with extreme data
The goal of this topic is to address issues of quantification of uncertainties and noise, of multi-scale, of multi-physics and extreme data. Mathematical methods, numerical analysis, algorithms and software engineering for extreme parallelism should be addressed. Novel and disruptive algorithmic strategies need to be explored to minimize data movement as well as the number of communication and synchronization instances in extreme computing. Parallel-in-time methods may be investigated to boost parallelism of simulation codes across a wide range of application domains.
Taking into account data-related uncertainties is essential for the acceptance of numerical simulation in decision making. A unified European VVUQ (Verification Validation and Uncertainty Quantification) package for Exascale computing should be developed. Access to the VVUQ techniques for the HPC community should be facilitated by providing software that is ready for deployment on supercomputers.
Exascale HPC ecosystem development
The ecosystem projects are small support projects. Two types of projects are in focus.
The first type should provide Coordination of the Exascale HPC strategy and International Collaboration. This include activities for promoting a joint community structuring and synchronisation and to further development and update of the Strategic Research Agenda for High Performance Computing as well as the application and applied mathematics exascale roadmaps. Projects should not only coordinate with projects and activities in FET, but also in the other parts of the Horizon 2020 programme.
The second type of projects should focus on Excellence in Exascale Computing Systems: The focus should be in boosting European HPC academic research excellence in future exascale-class computing cutting across all levels - hardware, architectures, programming, applications. Projects should include specific actions to better structure the European academic HPC research area, and create stronger links with HPC providers and HPC users, attract venture capital, promote entrepreneurship and foster industry take-up.
In the core Information and Communication Technologies (ICT) subprogramme, surprisingly little is done which directly relates to HPC. In the area of customised and low enery computing, there is a Research and Innovation action on "Programming environments and toolboxes for low energy and highly parallel computing". Considerable funding is available here.
The commission expects proposals in this area that will provide programming environments and tools optimised for specific application domains of significant economic value, covering the complete software stack from runtime systems to application programming.
Projects should support modern system architectures possibly including those based on heterogeneous processors while allowing for optimization of energy, performance, reliability, time predictability and system cost.
Model-based approaches and reuse and extension of existing platforms, libraries, frameworks and tools are encouraged. The results should be practically usable for application development for real-world use cases.
In total the Commission has 24 million euro available for R&D projects in this topic. Projects are expected to be around 4-6 million euro in size.
The European Commission promotes the I4MS (ICT Innovation for Manufacturing SMEs) initiative to support the European manufacturing industry through the adoption of ICT technologies. The ICT areas that are considered to be especially useful are:
For the coming 2 years, the EC is especially lookin at modelling, simulation and analytics: HPC Cloud-based modelling, simulation and analytics services with special emphasis on sustained service models; on providing real-time support; and on addressing comprehensively security and privacy issues at all levels.
So the focus is on mainstream HPC tied together with Cloud computing developments.
The Commission wants to fund at least one project in HPC Cloud-based simulation services. For all four topics together the budget is 32 million euro. So roughly about 8 million euro can be expected to be allocated to HPC Cloud-based simulation services.
There is also funding room for a small support project of around 250.000 euro.
The Human Brain Project is one of the two ICT related flagship projects supported by the European Commission. In the 2016-2017 timeframe a new sub project will be supported called "Interactive Computing e-infrastructure for the Human Brain Project FET Flagship". The proposal for this sub-project must support the Action Plan of the Human Brain project by making available a computing and memory system with an expected peak performance of 50 PetaFlop/s and 20 PetaByte of memory footprint as required for Human Brain Project simulations and for a period of about five years.
The components of the targeted system addressing interactive visualization and steering of large-scale brain simulations and their scalable integration are expected to comply with the capabilities and features defined during the Pre-Commercial Procurement (PCP) of technology readiness demonstration that was launched in the ramp-up phase of the HBP Flagship.
The Commission expects the proposal to explain how the novel features and capabilities of this infrastructure will also be made available to the pan-European High Performance Computing (HPC) Tier-0 infrastructure and services of PRACE, by guaranteeing at least another 15% of this new capacity to European researchers at large.
Because the supercomputing system will be especially targeted towards Human Brain simulation requirements, it will most likely also contribute to diversify the available leading-class HPC capabilities in the Tier-0 pan-European supercomputing infrastructure, and to the adoption and use in Europe of the most advanced HPC technology.
The Commission expects to spend about 25 million euro on this system.
The European Commission is putting considerable funding into HPC in the next two years. The main focus is on providing supercomputer services and developing hardware, tools, and techniques that are needed on the road to exascale.