6th Annual Conference of IT4Innovations and 1st Users Conference of IT4Innovations

31 Oct 2017, 08:30 → 1 Nov 2017, 16:00 Europe/Prague

atrium (IT4Innovations)

Branislav Jansik (IT4Innovations), Vít Vondrák (IT4Innovations)

The 6th Annual Conference of IT4Innovations is devoted to presentations from our research laboratories and those of our partners in the IT4Innovations Centre of Excellence. All IT4Innovations’ employees and employees of the project partners are welcome to attend the conference. The Annual Conference will take place at IT4Innovations on Tuesday 31 October from 8:30 untill 12:30.

After 12:30 on 31 October the 1st Users Conference of IT4Innovations will start. All our users as well as research and project partners from various organisations, research institutions and industry are welcome to attend and submit their research abstracts. In the afternoon keynotes and plenary talks will be given followed by a poster session. The Users conference will carry on on the next day, Wednesday 1 November, where attendees can discover more about our future upgrade plans, listen to keynotes by our prominent users and engage in discussions.

Types of contributions

Keynotes

Keynotes are invited talks by our prominent users which will be presented at selected times during the whole conference on October 31 as well November 1st.

Poster

Please note that the required poster size is A1 portrait orientation.

Each submitted abstract will be automatically accepted for poster session. If you do not wish to submit abstract for the plenary talks but would like to only bring a poster please fill in the abstract and choose poster from the drop down menu under Presentation type. All submitted posters will be accepted.

Users' contributions

The selected abstracts will be presented during the afternoon sessions on October 31st and  Users' contributions  on November 1. Each talk is expected to take 20 mins. This call is opened to all IT4Innovations infrastructure users.

Plenary talks

The selected abstracts will be presented in the Tuesday afternoon sessions. This call is dedicated to talks from the partners of the IT4Innovations Centre of Excellence project. Each talk is expected to take 20 minutes.

 

Leaflet Invitation

Slides

• A. Cammarata: Tailoring Nanoscale Friction in Transition Metal Dichalcogenides
• A. Miro: SoHPC Best Visualisation Award
• A. Vašatová: Software tool for cranial orthosis design
• B. Jansik: Presentation of IT4Innovations infrastructure and future plans
• B. Jansik: Users group
• J. Heyda: Thermodynamics of thermoresponsive polymer PNIPAM in aqueous solutions
• K. Carva: Calculating ultrafast laser-induced magnetization dynamics
• K. Kalousová: Transfer of heat and liquids through the high-pressure ice layer of Ganymede
• L. Říha, M. Merta, M. Jaroš: Selected Intel Xeon Phi Accelerated Libraries developed at IT4Innovations
• M. Čapek: Modelling of blood flow and thrombus formation
• M. Elbattah: SoHPC HPC Amabassador Award
• M. Komm: Particle-in-cell simulations of plasma-facing components for future thermonuclear reactors
• M. Lazecky: Automatized Sentinel-1 Monitoring System
• M. Pecha: PERMON in material engineering applications
• P. Stehlik: Examon Web: A Visualization Framework for Live and Collected Power, Energy, Performance and Operational Data in Supercomputers
• P. Veteška: What-If analysis with Floreon+
• S. Bohm: Generating discrete structures & HPC
• S. Kozmon: Using the IT4I infrastructure on the enzymatic reactions studies
• T. Martinovic: Recurrence quantitative analysis

Tuesday, 31 October

Registration of internal attendees

IT4Innovations in 2017: Vít Vondrák

Convener: Prof. Vít Vondrak (IT4Innovations, VSB-Technical University of Ostrava)

IT4Innovations Infrastructure: Branislav Jansík

Convener: Dr Branislav Jansik (IT4Innovations)

IT4Innovations research: Tomáš Kozubek

Summary of RP1: Jan Martinovič

Summary of RP2: Radim Blaheta

Summary of RP3: Tomáš Kozubek

10:45 Coffee Break

Summary of RP4: Jaromír Pištora

Summary of RP5: Miroslav Vozňák

Summary of RP6: Vilém Novák

Summary of RP7: Pavel Zemčík

Summary of RP8: Lukáš Sekanina

12:30 Lunch

Registration of external participants

Keynote 1: M. Komm: Particle-in-cell simulations of plasma-facing components for future thermonuclear reactors

1 Particle-in-cell simulations of plasma-facing components for future thermonuclear reactors

Power exhaust is one of the key problems, which need to be resolved in order to develop a safe, efficient and economically viable thermonuclear reactor. Heat fluxes impinging onto the plasma-facing components (PFCs) inside the reactor can easily exceed its material limits, especially if the real geometry of such components is taken into account.  We present results of particle-in-cell simulations, which were used to predict the heat flux distribution on the PFCs of tokamak ITER and to evaluate the role of sheath eletric fields.

Speaker: Dr Michael Komm (Institute of plasma physics of The Czech Academy of Sciences)

Slides Michael Komm: Particle-in-cell simulations of plasma-facing components for future thermonuclear reactors

Plenary talks 1: J. Heyda, M. Pecha, M. Lazecký

2 Thermodynamics of thermoresponsive polymer PNIPAM in aqueous solutions

Thermoresponsive polymers are very perspective materials, however, their utilization on theoretical grounds is currently limited, due to their unusual physico-chemical behavior, which possesses many challenges during the model development. In this contribution, we have developed an accurate parameterization of thermoresponsive polymer PNIPAM, which allows atomistic computer simulations in the broad temperature range. Most importantly, near the critical point, where the polymer is extremely sensitive to external stimuli. Large-scale, replica-exchange molecular dynamics simulations allowed to extract and analyze thermodynamic properties of the polymer in water and in aqueous solutions of salts. The results were compared with in-house calorimetry experiments. The interpretation is performed via the thermodynamic model, which connects the macroscopic thermodynamics with salt excess or depletion from polymer vicinity.

Speaker: Dr Jan Heyda (University of Chemistry and Technology, Prague)

Slides Jan Heyda: Thermodynamics of thermoresponsive polymer PNIPAM in aqueous solutions

3 PERMON in material engineering applications

Material science has an enormous impact on civil, automotive, and aircraft engineering, especially in the last two decades. Who does not know Elon Musk and his Tesla and SpaceX? Our story started three years ago, when we established cooperation with the Czech Technical University in Prague in development fatigue analysis software for HPC platform. Moreover, we have started research, in which material science is combined with machine learning approaches. If you want to hear more about this work, you are welcome to our presentation.

Speaker: Mr Marek Pecha (IT4Innovations & Dep. of Applied Math., VSB-Technical University of Ostrava)

Slides Marek Pecha: PERMON in material engineering applications

4 Automatized Sentinel-1 Monitoring System

This work represents an ongoing work at IT4Innovations related to satellite remote sensing topic. A preparation of database of pre-processed satellite synthetic aperture radar (SAR) images is a prerequisite for effective analyzes using SAR interferometry (InSAR), intensity and polarimetry analyzes. These techniques are able to observe especially vertical displacements in a very high sensitivity (reaching 1 mm/year) by InSAR and/or identify deforestation, flooded or burnt areas or prepare soil moisture maps from satellite radar which radiation penetrates clouds. Our database is connected to the Czech Collaborative Copernicus Ground Segment maintained by CESNET and contains bursts of Copernicus programme’s Sentinel-1 SAR satellite that have been preprocessed to the state of a consistent well-coregistered dataset by an open-source NASA/JPL software ISCE. A further processing time is significantly reduced in order to achieve so-called Persistent Scatterers (PS) or other, e.g. Small Baseline (SB) -based displacements velocity maps, as demonstrated using open-source STAMPS multitemporal InSAR processor. Our approach has several advantages over similar works - a processing speed and a preparedness for inclusion of various advanced techniques, for example a warning system against dangerous motion of landslides or infrastructure. Every new pre-processed burst can trigger a processing update that is able to detect unexpected changes in InSAR time series and therefore provide a signal for early warning against suspicious occurrence of a potential dangerous displacements. Using other open-source tools such as ESA’s SNAP we are able to deliver products such as soil moisture change maps or forest change maps over Czech areas. The work is ongoing, preliminary and expected results will be presented.

Speaker: Dr Milan Lazecky (IT4Innovations)

Slides Milan Lazecký: Automatized Sentinel-1 Monitoring System

15:00 Coffee Break

Keynote 2: K. Kalousová: Transfer of heat and liquids through the high-pressure ice layer of Ganymede

5 Transfer of heat and liquids through the high-pressure ice layer of Ganymede

The exploration of ocean worlds - planets or moons that harbor deep subsurface oceans - is prompted by the question of the emergence of life in places where liquid water is present. Measurements performed by the Galileo and Cassini missions confirmed the presence of these deep oceans within the interiors of some Jovian and Saturnian moons. A lot of attention is currently given to smaller moons (Europa and Enceladus) where the deep ocean is expected to be in a direct contact with the underlying silicate mantle – such conditions are similar to those at terrestrial sea floors where life develops. Ganymede, the largest moon in the solar system, possesses a larger amount of H2O so that a layer of high-pressure (HP) ice exists in its interior that seems to prevent the direct contact of water with silicates. We study the dynamics of this HP ice layer by solving the conservation equations of mass, momentum, and energy of a two-phase mixture by the finite element method. Our results indicate that melt can be produced at the silicate/HP ice interface and then transported through the layer by the upwelling plumes. Depending on the vigor of convection, it may stay liquid or freeze before melting again as the plume reaches the partially molten layer at the top boundary from where it is extracted into the ocean. Our results also suggest that these exchange processes were more likely earlier in Ganymede's history, when the HP ice layer was thinner.

Speaker: Dr Klára Kalousová (Charles University, Faculty of Mathematics and Physics, Department of Geophysics)

Slides Klára Kalousová: Transfer of heat and liquids through the high-pressure ice layer of Ganymede

Plenary talks 2: S. Böhm, S. Kozmon, P. Stehlík, K. Carva

6 Generating discrete structures & HPC

We are introducing Haydi -- a Python tool for generating discrete structures. It provides an easy way to define a structure from basic building blocks (e.g. Cartesian product, mappings) and then enumerate all elements, all non-isomorphic elements, or generate random elements. It is designed to provide a tool for checking claims on small instances and for generating "interesting" instances of a given class. The resulting program can be executed without modifications as a distributed application on a cluster. We have used Salomon for profiling and optimization testing. The project was developed as part of project GAČR 15-13784S.

Speaker: Dr Stanislav Böhm (IT4Innovations)

Slides Stanislav Bohm: Generating discrete structures & HPC

7 Using the IT4I infrastructure on the enzymatic reactions studies

Studying of the enzymatic reaction is still quite challenging task. Whole protein including the ligands and usually also the solvent should be used during the calculation. It brings together thousands or tens of thousands atoms. Such a big systems can bet threated only using the HPC resources and employing the specific methodologies as hybrid quantum mechanics / molecular mechanics (QM/MM) methods. Here the results of the enzymatic reaction mechanism study of the O-GlcNAc transferase (OGT) as a case study will be presented. The glycosyltransferase (OGT) post-translationally modifies a variety of proteins. The misregulation of O-GlcNAc-ylation is linked to a wide range of diseases, so knowing its reaction mechanism is very significant. Known OGT structures and experimental biochemical data suggest several possible mechanisms. In the present study, experimentally proposed mechanisms were investigated at the DFT QM/MM level. The sophisticated theoretical approaches such as hybrid QM/MM DFT Carr-Parinello ab initio molecular dynamics combined with a string method for reaction path optimization were used to investigate which of the proposed mechanisms might be the most probable.

Speaker: Dr Stanislav Kozmon (CEITEC MU, Masaryk University; National Centre for Biomolecular Research, Masaryk University; Slovak Academy of Sciences)

Slides Stanislav Kozmon: Using the IT4I infrastructure on the enzymatic reactions studies

8 Examon Web: A Visualization Framework for Live and Collected Power, Energy, Performance and Operational Data in Supercomputers

In the race toward exascale computing, supercomputing systems are getting more complex. This makes it more difficult to operate the computing resources, infrastructure and software components at the most efficient point. The first step to counteract these trends is to give users and system administrators a cockpit from which they can visually inspect status of the cluster and executed applications. We present Examon Web -- an open source framework for visualization of performance, power and energy statistics of HPC applications and cluster status. Examon Web is built on top of a fine grain monitoring framework which collects and handle a wide set of sensors and performance counters for the cluster computing resources, job scheduling data and infrastructure metrics all sampled at a fine granularity. Examon Web combines these information on a per job and per cluster base to provide visual insights on applications and cluster performance, power and energy.

Speaker: Mr Petr Stehlík (FIT VUT)

Slides Petr Stehlík: Examon Web: A Visualization Framework for Live and Collected Power, Energy, Performance and Operational Data in Supercomputers

9 Calculating ultrafast laser-induced magnetization dynamics

Since the first demonstration of femtosecond laser-induced demagnetization in 1996 [1], optical manipulation of magnetization has developed into a vibrant area of research that has explosively grown over the last decade. The most interesting one of discoveries is the demonstration of all-optical magnetization reversal and opto-magnetic recording with femtosecond laser pulses. All-optical magnetization switching (AOS) is emerging as a novel magnetic recording technology, its potential has been fully recognized by the magnetic recording industry, which enlisted AOS process on its roadmap towards ultra-high magnetic recording densities beyond 1 Tb/in^{2}. One of the critical problems in this field is related to the character of the transiently demagnetized state. Recently an insight into this problem has been provided by novel optical methods based on high harmonics generation [2]. Measurements of time-, energy-, and angle-resolved T-MOKE magnetic asymmetry spectra at M-edges have been compared to ab initio predictions for different deviations from the magnetic ground state, which allows us to draw conclusions about the importance of transversal magnetic excitations in the ultrafast laser-induced demagnetization [3]. An interesting magnetization dynamics has been observed in systems with 2 ferrimagnetically ordered sublattices: a complete reversal of magnetization in GdFeCo alloy. Most magnetic momentum in the system originates from Gd 4f states deep below the Fermi level, which prevents direct access to it by the laser. We calculate intraatomic exchange between Gd 4f and 5d orbitals [4] to allow mapping the problem to an effective orbital-resolved Heisenberg Hamiltonian. A subsequent simulation based on the LLG equation has reproduced the switching behavior and shown that the exchange coupling is sufficient to revert the large moment of Gd 4f shell on a ps time scale [5]. [1] E. Beaurepaire, J.-C. Merle, A. Daunois J.-Y. Bigot, Phys. Rev. Lett. 76, 4250 (1996). [2] Fan, T., et al., Proceedings of the National Academy of Sciences 112, pp. 14206–14211 (2015) [3] E. Turgut et al., Phys. Rev. B 94, 220408(R) (2016). [4] B. Frietsch et al., Nat. Commun. 6, 8262 (2015). [5] S. Wienholdt, D. Hinzke, K. Carva and others, Phys. Rev. B 88, 020406(R) (2013).

Speaker: Dr Karel Carva (Department of Condensed Matter Physics, Charles University)

Slides Karel Carva: Calculating ultrafast laser-induced magnetization dynamics

PRACE SoHPC Awards ceremony and presentations

slides Arnau Miró Mahmoud Elbattah

Poster session & Dinner

Wednesday, 1 November

Guided tour around the IT4Innovations infrastructure

Presentation of IT4Innovations infrastructure and future plans

slides Branislav Jansík: Infrastructure future plans

10:30 Coffee Break

Keynote 3: A. Cammarata: Tailoring Nanoscale Friction in Transition Metal Dichalcogenides

slides Antonio Cammarata: Tailoring Nanoscale Friction in Transition Metal Dichalcogenides

10 Tailoring Nanoscale Friction in Transition Metal Dichalcogenides

One of the main difficulties in understanding and predicting frictional response is the intrinsic complexity of highly non-equilibrium processes in any tribological contact, which include breaking and formation of multiple interatomic bonds between surfaces in relative motion. Moreover, under tribological conditions, local charge accumulation may take place and affect the tribological behaviour of the corresponding neutral structure. To understand the physical nature of the microscopic mechanism of friction and design new tribologic materials, we conducted a systematic quantum mechanic investigation at the atomic scale on prototipical Van der Waals MX2 (M=Mo, W; X=S, Se, Te) Transition Metal Dichalcogenides at different charge content. We combined the structural and dynamic information from group theoretical analysis and phonon band structure calculations with the characterisation of the electronic features using non-standard methods like orbital polarization and the recently formulated bond covalency, Normal-Mode Transition Approximation and cophonicity analyses. We formulated guidelines on how to engineer nanoscale friction, and finally applied them to design a new Ti-doped MoS2 phase. Thanks to the strong correlation between the electronic and the dynamical features of these systems, the present outcomes can be promptly used to finely tune other physical properties for the design of new materials with diverse applications beyond tribology. Ref. A. Cammarata, T. Polcar, Nanoscale 9 (2017) 11488 A. Cammarata, T. Polcar, Phys. Rev. B 96 (2017) 085406 A. Cammarata, T. Polcar, Phys. Chem. Chem. Phys. 18 (2016) 4807. A. Cammarata, T. Polcar, RSC Adv. 5 (2015) 106809. A. Cammarata, T. Polcar, Inorg. Chem. 54 (2015) 5739. A. Cammarata, J. Rondinelli, J. Chem. Phys. 141 (2014) 114704.

Speaker: Dr Antonio Cammarata (Czech Technical University in Prague)

Users Group

slides Branislav Jansík: Users group

13:00 Lunch

Keynote 4: L. Říha: Selected Intel Xeon Phi Accelerated Libraries developed at IT4Innovations

slides Lubomír Říha, Michal Merta, Milan Jaroš

11 Selected Intel Xeon Phi Accelerated Libraries developed at IT4Innovations

In this talk our developers will present three tools developed at IT4Innovations which support the Intel Xeon Phi accelerators installed at the Salomon supercomputer. ESPRESO library is a FEM package which uses the Xeon Phi to accelerate the linear FETI solver. ESPRESO contains new algorithms which can take advantage of the fast KNC memory to speed up the calculations. BEM4I library is a library of parallel boundary element based solvers which contains efficient vectorization techniques to fully utilize the potential of the KNC architecture and its vector units. Finally, BLENDER is an open source 3D modeling tool which contains its own Cycles render based on path tracing algorithm. We will present our new CyclesPhi renderer which adds multi-node parallelization and Xeon Phi acceleration into the original implementation.

Speakers: Dr Lubomir Riha (IT4Innovations), Dr Michal Merta (IT4Innovations), Mr Milan Jaros (IT4Innovations)

Slides Lubomír Říha, Michal Merta, Milan Jaroš: Selected Intel Xeon Phi Accelerated Libraries developed at IT4Innovations

Users' Contributions: T. Martinovič, M . Čapek, P. Veteška, A. Vašatová

12 Recurrence quantitative analysis

Recurrence quantitative analysis (RQA) is a quantification of structures in the recurrence plots. It is used to study the dynamical properties of the systems represented by the time series data. Although RQA is a powerful tool in time series analysis, it has an exponential computation complexity. Therefore, it is difficult to use RQA to analyze very long time series. This work focuses on the scalable parallel computation of the RQA. This algorithm greatly reduces spatial complexity of existing methods and allows parallelization with a small amount of communication. Scalability of the MPI (Message Passing Interface) implementation was tested and it is shown it scales very well.

Speaker: Tomáš Martinovič (IT4Innovations, VŠB - Technical University of Ostrava)

Slides Tomáš Martinovič: Recurrence quantitative analysis

13 Modelling of blood flow and thrombus formation

We will introduce the issue of computing the blood flow and blood coagulation process. We aim to model blood as a viscoelastic fluid. We want to capture the platelets and chemical species margination effect by red blood cells. The blood coagulation process is taken as growth of a phase field, which depends on both rheological and biochemical conditions in the vessel. We implement our model using FEM library deal.ii. In order to scale up to hundreds of cores we apply projection methods to solution of our coupled systems of equations. As the most promising projection method appears to be the incremental pressure correction splitting method, which we use for the Navier-Stokes part of our equations. Nonlinear terms in the Navier-Stokes equations are treated using the fixed point iteration. NS system is solved decoupled from the convection-diffusion-reaction transport equation for the blood coagulation species. As we are dealing with processes of different time scales, we use a specific adaptive time stepping scheme for the whole system of equations.

Speaker: Mr Marek Čapek (Mathematical Institute of Charles University)

Slides Marek Čapek: Modelling of blood flow and thrombus formation

14 What-If analysis with Floreon+

The Floreon+ (floods recognition on the net) is a system primary developed to provide flood monitoring and simulating capabilities that support decision process during crisis management within the Moravian-Silesian region in the Czech Republic. Because the system is developed with modular architecture, it allows for easy integration of new modules, or thematic areas to the system, such as traffic management or “on demand” What-If analysis, which are the latest features of the system. There are three main types of What-If analysis: Hydrological simulations, Spread of hazardous substances and Macro-modelling of traffic management simulations. All of these simulations are directly connected to the cluster, via HPC as a Service middleware. It allows computation of difficult simulations in real time for the specified area selected on the map. These types of simulations also allows detailed parameter specification for simulation of specific crisis situations. User is immediately informed about the progress of the calculation via the web interface. Obtained results are visualized through web interface on the map. What-If simulations are also prepared for a combination of different thematic areas together, such as automatic flood modeling with Macro-modeling of traffic management simulations.

Speaker: Mr Veteska Patrik (IT4I - ADAS)

Slides Patrik Veteška: What-If analysis with Floreon+

15 Software tool for cranial orthosis design

Cranial orthoses are used to correct an abnormal children head shape, and such they have to be designed individually. Customiza- tion of those orthoses is currently fully manual task. A software tool should make this process semi-automatic with only small intervention from the user and speed up the whole process. In the future, this tool will be part of process chain from 3D scanning of the patient head to the 3D printing of the final product. This will allow to produce the orthosis anywhere, without necessity to have expensive devices on one place. For high quality of 3D printing, 3D computer models with high- resolution meshes must be used. We start development of our tool by rapid testing of methodology. For this purpose we used open source soft- ware Blender. Although Blenders functions we used are more robust, they are also unnecessary computationally more expensive. For this rea- son we implemented the necessary transformation functions using radial basis functions (RBF) which can be easily modified to include rigid body movements.

Speaker: Ms Vasatova Alena (IT4Innovations)

Slides Alena Vašatová: Software tool for cranial orthosis design

Wrap up

Convener: Prof. Vít Vondrak (IT4Innovations, VSB-Technical University of Ostrava)