User experience: Excellent
User experience: Excellent!
The new CSC supercomputer Mahti was inaugurated into reserachers' use at a virtual seminar on 7 October. At the event, Erja Heikkinen, Deputy Director-General at the Ministry of Education and Culture, and Kimmo Koski, CEO at CSC underscored the importance of Mahti for Finland's academia. The vendors brought their greetings and six pilot users of Mahti shared their initial experiences of the computer. The seminar attracted an audience of nearly 400 viewers.
Pilot projects launched easily
CSC's Scientific Customer Panel selected 22 projects as the Mahti Pilot Users. Projects of very different types were chosen as pilots. They were expected to produce a heavy load on the new supercomputer and to stress test the new system in a versatile manner.
Miguel Caro (Aalto University) uses machine learning as an aid in the modeling of interatomic potential. Targets of his research include a mechanism for growing carbon nanotubes and finding a catalyst for turning carbon dioxide into synthetic fuel.
– It is computationally very expensive to calculate quantum mechanical descriptions of interaction between atoms, and classical methods are not sufficiently accurate. With the help of machine learning we can improve this situation. We benefit from having Mahti's CPU calculations in quantum mechanics and Puhti in training artificial intelligence, which is one of the bottlenecks.
– We thank the Academy of Finland for its funding and CSC for CPU time and for human time, which is more valuable than CPU time. CSC has world-class resources, and world-class user support. Requesting computing time is very simple, and the time saved can be used for new science.
Vivek Sharma (University of Helsinki) applies high performance computing and molecular dynamics to biology. Many well-known structures are static snapshots, but biomolecules are in constant movement when they do their work. Sharma's goal is to add dynamics into static structures. A good example is the folding of proteins, in which understanding movement is very important.
Sharma is researching an especially important enzyme in our bodies, the respiratory complex I. This enzyme has a very central role in cellular respiration. Cellular respiration is energy metabolism in cells, in which cells release energy contained in nutrition so they can use it.
– The computational approach is necessary for us to understand the operations of biomolecules and for this we need a supercomputer.
In his pilot, Sharma calculated the atomistic molecular dynamic simulation of a microsecond for the 1.3 million atom model out of the entire complex.
– Puhti and Mahti are ideal for these types of large-scale simulations.
Andrea Sand (University of Helsinki) is researching the effects of radiation on materials, and she discusses her own project Electronic Excitations in Tungsten (EETT).
Quantum mechanics is needed in the study of electrons, which requires both heavy computing and huge resources.
The method used by Sand uses one core per electron.
– In this work I computed a system of 1000 atoms and 12 electrons per atom. I was able to easily divide the work to 12,000 cores, and that is what Mahti makes possible.
Maarit Käpylä (Aalto University) studies the magnetic fields of the Sun and shows how a small-scale dynamo takes place in the sun and how it affects space weather – and how this was computed using CSC's “Mighty Mahti”.
The Sun is a giant ball of gas, but it is also a huge magnet, and magnetic activities of different kinds cause eruptions of particles from the Sun. These eruptions can be seen on Earth as the Aurora Borealis, but they also affect space weather and disrupt the operations of satellites, electricity grids, communications, and aviation, among other things. For this reason, it is important to understand how Solar activity drives the space weather.
– The solar dynamo is a phenomenon that is not well-known. The surface of the Sun boils, there is perpetual turbulence there and numerical modeling is the only way to study it. Thanks to the resources of this Mahti pilot, we were able to drive a larger model, and the result that we got was that small-scale dynamo activity should be possible in the Sun. A bigger model also produces considerably more data, and we need to prepare for it, especially in the upcoming exascale supercomputers.
Rupert Gladstone (Arctic Centre, University of Lapland) simulates the uncertain future of the Antarctic ice sheet and its impact on the rise in the sea level. He primarily uses the Elmer/Ice software developed at CSC.
Gladstone has worked in many countries and with many HPC centers and has never worked with a company that has so clear linkage between HPC and science.
– My partners at CSC do not only operate HPC computers – they are actively developing scientific codes and engaging in science themselves. I think that this is beneficial.
With Puhti and Mahti, researchers can use increasingly precise models with better resolution and combined ice sheet – ocean models.
– With the resources that are now available to us we can already achieve remarkable improvements in ice sheet modeling. Our greatest challenge now is to develop the software into something that would allow us to take full advantage of the existing equipment.
Karoliina Honkala (University of Jyväskylä) is testing the effectiveness of Mahti by using molecular dynamic simulation based on density functional theory. She is studying rhodium and platinum clusters on a zirconium surface that can easily be used as catalysts.
– The goal is to understand how many structurally different kinds of cluster isomers exist for each metal and each cluster size. This became possible with the help of the computational capacity of Mahti. Without it, this could not have been calculated.
– User experience: excellent! We were able to get started as soon as Mahti was turned on. There were hardly any problems at all.
Researchers involved already in early stages of planning procurement
Erja Heikkinen says that users were included already in the early stages, as soon as planning for the procurement began. In a questionnaire on what was needed, researchers brought up data management and artificial intelligence, which are rising in importance, to a level commensurate with HPC. One great change that came with the DL2021 environment was when CSC's clientele expanded to include the research institutes of other ministries.
– The new environment made it possible to react rapidly to a fast-emerging need for research caused by Covid-19. This fast track was an encouraging example of broader societal effectiveness, Erja Heikkinen said.
In his speech Kimmo Koski raises the transition from computing to data computing, spiced with artificial intelligence. The aim is to build an ecosystem around Mahti.
– Mahti has been enabling science efficiently from day one. The utilization ratio has been as high as 99%.
Check the Mahti inauguration seminar on the CSC YouTube channel
Read more about Vivek Sharma's research
Read more about Rupert Gladstone's research
Read more about Miguel Caro's research