In Last Name Alphabetical Order:
- Haohuan Fu (Tsinghua University, CN)
- Sadakazu Haino (Academia Sinica, TW)
- David Jones (UCL, UK)
- Erwin Laure (KTH, SE)
- Davide Salomoni (INFN, IT)
- Matthias Wolf (OIST, JP)
Tsinghua University, CN
Non-Linear Earthquake Simulation on Sunway TaihuLight
The Sunway TaihuLight supercomputer is the world's first system with a peak performance greater than 100 PFlops, and a parallel scale of over 10 million cores. Different from other existing heterogeneous supercomputers, the system adopts its unique design strategies in both the architecture of its 260-core Shenwei CPU and its way of integrating 40,960 such CPUs as 40 powerful cabinets. This talk would first introduce and discuss design philosophy about the approach to integrate these 10 million cores, at both the processor and the system level. Based on such a system design, we adopt comprehensive memory-related optimizations to resolve TaihuLight's bandwidth constraints, and propose on-the-fly compression, which doubles the maximum problem size and further improves the performance by 24%. The resulting design achieves 15% of the peak, using 160,000 MPI processes, 10,400,000 cores, and enables Simulation for 18-Hz and 8-meter scenarios.
Haohuan Fu is the deputy director of the National Supercomputing Center in Wuxi, managing the research and development division. He is also an associate professor in the Ministry of Education Key Laboratory for Earth System Modeling, and Department of Earth System Science in Tsinghua University, where he leads the research group of High Performance Geo-Computing (HPGC). Fu has a PhD in computing from Imperial College London. Since joining Tsinghua in 2011, Dr. Fu has been working towards the goal of providing both the most efficient simulation platforms and the most intelligent data management and analysis platforms for geoscience applications. His research has led to led to two consecutive ACM Gordon Bell Prizes (fully-implicit atmospheric dynamic solver in 2016, and non-linear earthquake simulation in 2017), Significant Papers of FPL (27 out of 1,765 publications in 25 years of FPL), and Best Paper Award (3 out of 278 submissions) of ICTAI 2015.
Institute of Physics, Academia Sinica
Challenging Einstein's Theory of General Relativity by Gravitational Waves with Advanced Computing Technologies
The recent historic discovery of Gravitational Waves (GW) by LIGO won the Nobel Prize in physics in 2017 and opened a new era of GW astronomy. After about 100 years since the completion of General Relativity by Einstein, for the first time we can test this theory at extreme gravity conditions by using GW signals. The advanced computing technologies such as GRID and GPU play important roles to achieve these goals. The summary of world-wide GW detection network and data analysis approaches will be discussed.
S. Haino got his Ph.D at University of Tokyo in 2004 and has been working in experimental particle physics research in space such as Alpha Magnetic Spectrometer (AMS). He joined KAGRA (Large Cryogenic Gravitational-wave Telescope), and has been serving as a chief of KAGRA calibration subsystem and KAGRA Scientific Congress (KSC) board member for foreign institutes as well as Executive for International Collaboration at GSROC (The Gravitational Society of the Republic of China). He won Academia Sinica Research Award for Junior Research Investigators in 2015 and Academia Sinica Career Development Award in 2016 and served as a member of International Scientific Program Committee (ISPC) member of the 35th International Cosmic-Ray Conference (ICRC).
Applying deep learning to the prediction of protein structure and function
In my talk I will be reviewing some new applications of deep learning in bioinformatics, illustrated by methods that have recently been developed in my laboratory. I will focus on two fundemental challenges in modern biology: the prediction of gene function from both sequence and "multi-omic" data sets, and the prediction of protein 3-D structure from amino acid sequence. The techniques we have employed include network or graph embedding, deep fully convolutional networks and multitask deep learning.
David Jones received his B.Sc. in Physics from Imperial College and then went on to do an M.Sc. in Biochemistry at Kings College London, followed by a PhD in Computational Biology at University College London. After completing a Wellcome Trust Biomathematics Fellowship at UCL, in 1995 he was awarded a Royal Society University Research Fellowship to set up his own lab at the University of Warwick. In 1999, at 32 years of age, he became the first Professor of Bioinformatics in the UK at Brunel University. In 2001, he was appointed Professor of Bioinformatics at University College London (a joint appointment between the departments of Biochemistry and Computer Science). He is currently the Director of the Bloomsbury Centre for Bioinformatics, a joint research centre between UCL and Birkbeck College. His lab aims to develop and apply state-of-the-art mathematical and computer science techniques to problems now arising in the life sciences, particularly those driven by the post-genomic era. David's main research interests include protein structure prediction and analysis, simulations of protein folding, applications of Hidden Markov Models, transmembrane protein analysis, machine learning applications in bioinformatics, biological text mining, de novo protein design methodology, and genome analysis including the application of intelligent software agents. He is also the author of a number of very well-known bioinformatics applications: THREADER, GenTHREADER, PSIPRED and MEMSAT, and was one of the original co-authors of the CATH protein structure classification scheme (along with Profs. Christine Orengo and Janet Thornton). David was also a co-founder of Inpharmatica Ltd., which was founded in 1998 as a spin-out company from University College London and acquired by Galapagos NV in 2007. The company used a combination of bioinformatics and chemoinformatics to look at the relationships between the structure and function of proteins, and the binding of chemical groups to these proteins leading to the discovery of novel drugs.
Preparing Applications for the New Era of Computing
We are entering a new Era of computing with large scale HPC systems soon reaching performance levels of Exaflops and large scale cloud infrastructures providing unprecedented levels of compute power. These greatly enhanced computational capabilities will enable new science, pushing both the boundaries of existing computational science and enhancing new domains like artificial intelligence and high performance data analytics.
These enormous advances are being enabled by new hardware technologies, particularly changing CPU and memory design. Unfortunately, these changes are not transparent to applications and call for revised algorithms and implementations; appropriate programming environments, workflows, and data management systems; and a deep understanding of the hardware developments to come.
Dedicated effort have started in Europe to address these issues and in this talk we review some of the key challenges and present Europe's approach to them. Highlights of some selected projects, like the BioExcel Centre of Excellence for Computational Biomolecular Research will also be presented.
Prof. Erwin Laure (male) is Professor for High Performance Computing at KTH Stockholm. He leads the Department of Computational Science and Technology and is Director of the PDC - Center for High Performance Computing. Prior to this position he was the Technical Director of the "Enabling Grids for E-Science in Europe
(EGEE)" project working at CERN. He has more than 20 years experiences in High Performance Computing, is the Coordinator of several leading European Exascale projects (e.g. “EpiGRAM”, “ExaFLOW”, and “BioExcel”), and actively involved in major e-infrastructure projects (PRACE, EUDAT). His research interests include programming environments, languages, compilers and runtime systems for parallel and distributed computing, with a focus on exascale computing.
The European Open Cloud for e-Science towards automation, service composition, big data analytics and new frontiers in data management
The need to reduce fragmentation in the discovery, access, effective and efficient use of distributed resources in Europe and elsewhere is a well-known problem. This is one of the issues targeted, among others, by the European Open Science Cloud (EOSC) initiative. In this keynote, a description of the main problems will be given, together with a summary of the key technological solutions developed by the INDIGO-DataCloud project. These solutions will find place in EOSC-hub, a Horizon2020 project jointly prepared by INDIGO-DataCloud, EGI and EUDAT, running from 1/1/2018 to 31/12/2020, involving more than 100 international partners and providing an integration and management system of the European Open Science Cloud. The keynote will also discuss technological developments foreseen in two recently started Horizon2020 projects, called eXtreme-DataCloud and DEEP-Hybrid DataCloud, that will deliver exciting new services evolving several INDIGO-DataCloud components in key areas, such as intelligent dataset distribution, data lifecycle management, flexible metadata management for big data sets, real-time and streaming-based data ingestion and processing, and several others.
Davide Salomoni is Director of Technology at the Italian National Institute for Nuclear Physics (INFN). He has 27 years of international experience in private and public environments related to distributed computing and communication technologies. He currently leads the Software Development and Distributed Systems department at CNAF, the INFN National Center dedicated to research and development on IT technologies, located in Bologna, Italy.
He was Project Coordinator of INDIGO-DataCloud, a 26-partners project funded with 11M€ by the EC Horizon2020 framework program targeted to multi-disciplinary scientific communities and to resource providers. INDIGO-DataCloud developed innovative open source computing and storage solutions, deployable on heterogeneous, hybrid distributed infrastructures. He was also part of the core team that prepared the EOSC-hub proposal, a project running from 1/1/2018 to 31/12/2020 led by EGI, EUDAT and INDIGO-DataCloud, involving more than 100 international partners, that will provide an integration and management system of the European Open Science Cloud.
He leads or participates to several other national and international projects and advisory groups on distributed architectures, is the coordinator of the INFN Cloud Computing national working group, member of the INFN Scientific Computing Committee and is engaged with activities and collaborations with Universities, Public Administrations, research institutions and commercial companies through seminars, courses, lectures and joint programs.
From 2005 to 2011 he was in charge of designing, implementing, starting up and managing the large computing farm installed at the INFN National Computing Center located at CNAF, Bologna, which currently counts about 18,000 CPU cores, 25 Petabytes of disk storage and 50 Petabytes of tape storage.
From 2003 to 2005 he had the role of senior scientist at NIKHEF (Nationaal instituut voor subatomaire fysica, the Dutch institute for research in Astro-Particle Physics) in Amsterdam, The Netherlands. He focused there on distributed computing and in particular on Grid computing technologies, contributing to the management and development of the Dutch national computing center at NIKHEF/SARA.
From 2001 to 2003 he was Technical Manager of the Dutch Internet Team at COLT Telecom in Amsterdam, The Netherlands. He contributed there to the design and operational readiness of the COLT NL data center and to the definition and implementation of several commercial products involving data and IT managed services. He was also a member of the architectural group of the COLT Europe Internet Division.
From 1999 to 2001 he worked at the Stanford Linear Accelerator Center (SLAC) in Menlo Park, USA. He chaired the SLAC networking group, focusing in particular on the optimization of the local networks, on network data transfer for physics experiments and on the evolution of wide-area high-speed data transmission technologies.
From 1991 to 1998, after his MD in Physics from the University of Bologna (1990), he worked at INFN CNAF on networking technologies, designing and developing distributed monitoring systems for the then-current DECnet, IP and X.25 networks used by INFN and by other Italian scientific communities. He represented INFN in several national and international working groups linked to the design and development of network communication protocols and distributed infrastructures. He was one of the designers and implementers of the Italian academic and research network (GARR), and was the architect and first manager of its Network Operations Center.
Okinawa Institute of Science and Technology Graduate University (OIST)
Grid computing and cryo-EM
Throughout the last 25 years, single particle cryo-electron microscopy (cryo-EM) has continuously evolved into a powerful modality for determining the 3D structure of radiation-sensitive biological macromolecules, culminating in the award of the recent Nobel prices in Chemistry 2017. This development has been enabled by constant maturation of image processing algorithms in concert with the emergence of direct electron detectors, improvements in electron optics, stage stability and microscope automation, resulting in ever-growing image data volumes. I will give a brief overview of the state-of-the-art single particle cryo-EM workflow and the process of 3D reconstruction from images. Although the data volume generated by a single modern cryo electron microscope is still far from what detectors in particle physics produce each day, the growing interest and combined word-wide deployment of these facilities have already created bottlenecks in data storage and processing capacity. I will outline the resulting challenges faced by cryo-EM labs, describe the current modus operandi and a road map to grid computing in this field, which may enable institutions without expensive local infrastructure to benefit from distributed compute resources.
Dr. Wolf joined OIST as Assistant Professor in 2012, where he is leading the Molecular Cryo-Electron Microscopy Unit. He was a postdoctoral research fellow at Harvard Medical School in the laboratory of Prof. Steven Harrison with an emphasis on molecular virology. Matthias received a PhD in Biophysics from Brandeis University/MA working with Prof. Nikolaus Grigorieff on the structure of viruses and ion channels using cryo-EM. Previously, he studied pharmaceutical chemistry with an emphasis on computer-aided drug design and holds a Master's degree in Pharmacy from the University of Innsbruck, Austria.