LUMI tulee vuoden päästä
LUMI supercomputer will be placed in CSC's data center in Kajaani, Finland. Operations will start in Q1/2021.
LUMI will be here in one year
Supercomputers are required in top research in a wide range of data- and computing-intensive fields, for example in material science, fusion research, climate and pharmaceutical research. There are a lot of research problems which can't be solved without massive computing resources, and this type of research would be impossible without such computing resources.
During the time of writing this text, there were only two European-based supercomputers in the top ten list of world's fastest computers: the supercomputer named Piz Daint in Switzerland and the SuperMUC-NG supercomputer in Germany.
The EuroHPC Joint Undertaking (JU) is pooling European resources to develop top-of-the-range exascale supercomputers for processing big data, based on competitive European technology with the support of European Union. With the help of EuroHPC, a world-class computing and data infrastructure will be created to Europe. The budget for the EuroHPC JU in 2019–2020 is over 960 million Euros, which includes public investments from the EU and participating states as well as investments from private partners.
The first aim of EuroHPC is to purchase three pre-exascale supercomputers for Europe by the year 2021. One exaflop means processor computing power corresponding to 1018 floating point calculations per second or one trillion calculations per second. The aim of EuroHPC is to develop European HPC expertise and technology so that in 2022–2024, one of the fastest exascale supercomputers of the world, based on European technology, would be purchased for Europe. A total of 30 European countries have already signed the EuroHPC Declaration.
In June 2019, the EuroHPC Joint Undertaking decided to place the EuroHPC pre-exascale supercomputers to Finland, Italy and Spain. The LUMI (Large Unified Modern Infrastructure) consortium was formed for the EuroHPC application process. The countries involved in the LUMI Consortium are Finland, Belgium, Czech Republic, Denmark, Estonia, Norway, Poland, Sweden and Switzerland.
Sustainable data center ecosystem in Kajaani
LUMI will be placed to CSC's data center in Kajaani, Finland where CSC has operated a data center since 2012. The data center is located in an old paper mill, providing modern data center facilities and a safe environment.
The benefits of the old paper mill site include the abundant availability of energy up to 200 megawatts. Few, if any other HPC centers in the world have similar facilities. In addition, the waste heat generated by the supercomputers will be reused in the local district heating network. This decreases the costs and reduces the carbon footprint. The energy used in CSC's data center is generated 100% with hydropower, so the carbon footprint of the data center is actually negative. The reliable and fast backbone connections of the data center are designed for the purposes of HPC.
An interesting curiosity is that the waste heat generated by LUMI can replace heat produced by fossil fuels. The waste heat of LUMI to be used in Kajaani's district heating network equals up to 20 per cent of Kajaani's yearly need of energy used in the area's district heating. The reuse of waste heat reduces the carbon footprint of Kajaani by amount that equals the use of 4000 passenger cars yearly.
One of the most competitive supercomputers in the world
When the operations of LUMI will start in the beginning of 2021, LUMI will be one of the most competitive supercomputers in the world. The design philosophy for LUMI was to create a platform which make the use of AI, especially deep learning, and traditional large scale simulations combined with massive scale data analytics possible in solving one research problem.
The theoretical peak performance of LUMI will be over 200 petaflops (2*1017 floating point operations per second), which as well as the application performance are approximately ten times faster compared to the Piz Daint supercomputer. LUMI achieves its high performance with a large number of nodes with accelerators (GPUs). In addition, the system is complemented by a CPU only partition, IaaS cloud services and a large object storage solution. LUMI will have over 60 petabytes of storage with a sizeable flash layer providing more than 1 terabyte/s of bandwidth.
This is the first time ever when nine countries are investing in one HPC system. 24 per cent of LUMI's computing capacity will be directly used for the benefit of Finnish researchers and society. In addition, Finnish researchers can apply for computing capacity via peer review process from the two other EuroHPC pre-exascale supercomputers.
Investing in a supercomputing infrastructure alone does not guarantee Europe leadership in innovation and science. This is why EuroHPC is focusing also in HPC research and development, training and HPC competence development. The end users will see this as EuroHPC Competence Centers which will provide training events, HPC consultation and provide help with software modernization projects. LUMI infrastructure's distributed user support (LUST, LUMI User Support Team) brings Finnish LUMI users competence and know-how which CSC has not been able to earlier provide alone.
The operations of LUMI are foreseen to start already in one year, in early 2021. The user communities should start to think already now, how to utilize the significant increase of computing capacity, and start to prepare software for LUMI's architecture. Software using only conventional x86 processors can be modernized to employ GPUs, or alternatively research groups may start to use software already using GPUs with similar functionalities.
Added value for Finnish research community
LUMI will provide world-class resources for Finnish researchers: this will increase the possibilities to solve difficult research problems on a level that has been out-of-reach with CSC's national computing resources. The computing resources of LUMI are a significant benefit for Finnish research groups in international research projects, and in addition it will help researchers to get in to different international collaboration projects. The access to LUMI resources will be a remarkable bonus when recruiting researchers and specialists to Finland, as LUMI will be one of the best known scientific instruments during its lifespan.
The research community will eventually decide what kind of problems will be solved with LUMI. We foresee LUMI to help remarkably for example in with following kind of research questions:
• More precise climate models and the interconnection of different climate models: how will the living conditions change when the climate is warming?
• The sequencing and analyzing of full genome combined with data analysis and correlations to clinical data will bring more light to diseases and heritary diseases: causes and personalized treatment and medicine
• Artificial intelligence (deep learning): analyzing large data sets (simulated and measured) and reanalyzing e.g. in atmospheric science, environmental science, climate modelling, material science and language science
• Self-driving cars and vessels: the study of algorithms related to these with previously unprecedented computing power
• Social sciences: large scale data set analytics from social networks and the modelling of different phenomena
• Facts: LUMI supercomputer
• At the time of installation, one of the world's fastest computer systems. Performance more than tenfold compared to Europe's fastest supercomputer today (Piz Daint, Switzerland). LUMI is the size of a basketball court.
• The theoretical computing power of over 200 petaflops. LUMI's computing power is equivalent to the combined performance to 600,000 of the latest MacBook Pro computers. The supercomputer achieves its high performance with a large number of nodes with accelerators (GPUs). In addition, the system is complemented by a CPU only partition, IaaS cloud services and a large object storage solution.
• Storage: over 60 petabytes with a sizeable flash layer providing more than 1 terabyte/s of bandwidth.
• In use Q1/2021.
The article has previously been published in Arkhimedes magazine 3-4/2019 (in Finnish).
comments powered by Disqus