1 Domain-oriented services and resources of Polish Infrastructure for Supporting Computational Science in the European Research Space Progress and Future Plans for PL-Grid Development J. Kitowski, K. Wiatr, Ł. Dutka, T. Szepieniec, M. Sterzel and R. Pająk ACK Cyfronet AGH, Kraków, Poland PL-Grid Consortium KU KDM 2014, Zakopane, 12-14.3.2014
Outline 2 Motivation and Background Consortium Family of PL-Grid Projects Domain-specific solutions and services Overview Use-cases Future Plans Conclusions
Motivation and Background 3 Addressing Computational Science Problems PL-Grid Consortium (2007) 5 Polish Computer Centres Continuous development based on: Projects funded by the European Regional Development Fund as part of the Innovative Economy Program Polish Optical Internet (PIONIER) (2x10++ Gb/s) Close international collaboration EGEE, EGI InSPIRE, EMI, PRACE ESFRI projects, e.g., CTA, EPOS, ERIC cases..) RoadMaps for e-infrastructure developments Previous / Current projects (5FP, 6FP, 7FP, EDA ) Integration with international platforms
Family of PL-Grid Projects coordinated by Cyfronet 5 PL-Grid Project (2009 2012) (run by PL-Grid Consortium) Budget: total 21 M, from EU 17M Number of people involved: ca. 80 (total, from different Polish Centres) Outcome: Common base infrastructure National Grid Infrastructure (NGI_PL) Resources: 230 Tflops, 3.6 PB PLGrid PLUS Project (2011 2015) (run by PL-Grid Consortium) Budget: total ca.18 M, from EU: ca.15 M Number of people involved: ca. 120 IT scientists Computational scientists Field experts Close collaboration between Partners Homogeneous project view Expected outcome: Focus on users Domain specific solutions: 13 domains (Specific computing environments) QoS by SLM Extension of resources and services by: 500 Tflops, 4.4 PB Keeping diversity for users Clusters (thin and thick nodes, GPU), SMP, vsmp, Clouds
Family of PL-Grid Projects coordinated by Cyfronet, con t. 6 PLGrid NG Project (2014 2015) (run by PL-Grid Consortium) Budget: total ca. 3,6 M Expected outcome: Optimization of resources usage, training Increase of scientific level Extension of domain specific solutions by 14 add l domains Extension of resources and services by: ca. 8 Tflops, some PB
Focus on Users 7 Computer centres Hardware/Software User friendly Services Help Desk Domain Experts QoS/SLM Grants Real Users
TOP500 June 2011 Nov. 2012 June 2013 Nov. 2013 Polish Sites 8 Rank Site System Cores Rmax (TFlop/s) Rpeak (TFlop/s) 81 106 113 145 Cyfronet Poland Zeus - Cluster Platform SL390/BL2x220, Xeon X5650 6C, 2.660GHz, Infiniband QDR, NVIDIA 2090 Hewlett-Packard 11,694 23,932 25,468 25,468 104.8 234.3 266.9 266.9 124.4 357.5 373.9 373.9 143 170 221 ICM Warsaw Poland BlueGene/Q, Power BQC 16C 1.600GHz, Custom Interconnect IBM 16,384 16,384 16,384 172.7 189.0 189.0 209.7 209.7 209.7 163 TASK Gdańsk GALERA PLUS -- Action Xeon HP BL 2x220/BL490 E5345/L5640 Infiniband ACTION 10,384 65.6 97.8 194 WCSS Wroclaw Cluster Platform 3000 BL2x220, X56xx, 2.66 GHz, Infiniband Hewlett-Packard 6,348 57.4 67.5 375 PCSS Poland Rackable C1103-G15, Opteron 6234 12C 2.40 GHz, Infiniband QDR SGI 9,498 89.8 211.1
Cyfronet at TOP500 list 9 List Rank System Vendors Cores 11/2013 145 06/2013 113 11/2012 106 06/2012 89 11/2011 88 06/2011 81 11/2010 85 06/2010 161 11/2008 311 11/1996 408 SPP1600/XA-32 06/1996 408 SPP1200/XA-32 Zeus - Cluster Platform SL390/BL2x220, Xeon X5650 6C 2.660GHz, Infiniband QDR, NVIDIA 2090 Zeus - Cluster Platform SL390/BL2x220, Xeon X5650 6C 2.660GHz, Infiniband QDR, NVIDIA 2090 Zeus - Cluster Platform SL390/BL2x220, Xeon X5650 6C 2.660GHz, Infiniband QDR, NVIDIA 2090 Zeus - Cluster Platform SL390/BL2x220, Xeon X5650 6C 2.660GHz, Infiniband QDR, NVIDIA 2050/2090 Zeus - Cluster Platform 3000 BL 2x220, Xeon X5650 6C 2.66 GHz, Infiniband Zeus - Cluster Platform 3000 BL2x220, L56xx 2.26 Ghz, Infiniband Zeus - Cluster Platform 3000 BL2x220, L56xx 2.26 Ghz, Infiniband Cluster Platform 3000 BL2x220, L56xx 2.26 Ghz, Infiniband Zeus - Cluster Platform 3000 BL2x220, L54xx 2.5 Ghz, Infiniband Rmax (GFlop/s) Rpeak (GFlop/s) Hewlett-Packard 25932 266900 373900 Hewlett-Packard 25932 266900 373900 Hewlett-Packard 23932 234000 357500 Hewlett-Packard 13944 185316 271113 Hewlett-Packard 15264 128790 162409.0 Hewlett-Packard 11694 104765.1 124424.2 Hewlett-Packard 9840 88050.7 104697.6 Hewlett-Packard 6144 39934.5 55541.8 Hewlett-Packard 2048 16179 20480 Hewlett-Packard (Convex) Hewlett-Packard (Convex) 32 5.5 7.7 32 4.0 7.7
Supercomputer Zeus 10 Xeon, 23 TB, 169 TFlops Opteron, 26 TB, 61 TFlops Xeon, 3,6 TB, 136 TFlops Xeon, 6 TB, 8 TFlops ZEUS Statistics 2012 (2013) Users needs taken into account Almost 8 mln jobs 21,000+ daily 80 mln CPU hours 9130 years 800+ users (ca. 2000) 100PB+ usage of scratch (350 PB) The longest job: 76 days The biggest job: 576 cores (1024) Ca. 50% CPU time for multicore jobs
PLGrid Plus: Activities in general 12 Integration Services National and International levels Dedicated Portals and Environments Unification of distributed Databases Virtual Laboratories Remote Visualization Service value = utility + warranty SLA management Computing Intensive Solutions Specific Computing Environments Adoption of suitable algorithms and solutions Workflows Cloud computing Porting Scientific Packages Data Intensive Computing Access to distributed Scientific Databases Homogeneous access to distributed data Big Data Data discovery, process, visualization, validation. 4th Paradigm of scientific research Instruments in Grid Remote Transparent Access to instruments Sensor networks Organizational Organizational backbone Professional support for specific disciplines and topics Rich international collaboration (EGI.eu, EGI INSPIRE, EMI, PRACE..
PLGrid Plus: Domain Grids Program for strategic science domains and important topics of Polish/European Science Access to the software packages is provided by: glite Unicore QCG (usage 2011-2013: ca. 3000 cores/month) (usage 2013: ca. 60-100 cores/month) (usage H2 2013: ca. 2300 cores/month) local Already identified 13 communities/scientific topics: 13 Astrophysics High Energy Physics Life Sciences Quantum Chemistry and Molecular Physics Synchrotron Radiation Energy Sector Metallurgy Nanotechnology Acoustics Ecology Bioinformatics Health Material Science
PLGrid NG New Generation Activities in general 14 Additional/new groups of experts involved Security on new applications, audits In the development stage Before deployment During exploitation Optimization of resource usage -- IT experts Operation Center Optimization of application porting User support First-line support Helpdesk Domain experts Training
PLGrid NG: Domain Grids 15 Identified 14 communities/scientific topics: Medicine składowanie, bezpieczeństwo i analiza danych, diagnostyka OpenOxides dane i badanie struktur tlenkowych, spintronika Mathematics data mining, data analytics, analiza statystyczna także dla potrzeb zarządzania zasobami Biology analiza i wykorzystanie danych NGS, integracja aplikacji z systemem Kepler Hydrology wsparcie prognoz hydrologicznych przez prognozy pogody w wysokiej rozdzielczości Geoinformatics integracja danych przestrzennych Meteorology prognozy pogody i jakości powietrza w wysokiej rozdzielczości Complex Networks inteligencja obliczeniowa, ekstrakcja wiedzy, udostępnienie danych (referencyjnych) ebaltic-grid modelowanie lodu i oceanu dla Bałtyku UNRES gruboziarnista symulacja dynamiki i zwijania białek w skali ms, z pomocą MD, MM, MC VPH medycyna spersonalizowana, biologia molekularna, bio-banki nowej generacji Computational Chemistry integracja danych eksperymentalnych, zaawansowane metody obliczeniowe Nuclear Power wypalenie paliwa (MC), symulacja maszyn rotodynamicznych Metal Processing modelowanie, badania wariantowe
16 Examples of domain specific solutions and services (PLGrid PLUS use cases)
Deployed domain specific services Acoustics 18 Project Database the service is a web portal designed to provide PL-Grid users with the numerical database (simulation results) and experimental database (measurements results). Noise maps the service is an application used to generate noise maps in urban environments, based on data provided by the user. Animation: calculated dynamic noise map
Acoustics Noise Map service 19 The Noise Map service for creating maps of noise threats for roads, railways and industrial sources. Integration of the software service with the network of distributed sensors brings a possibility of making automatic updates of noise maps for a specified time period. Illustration of an application of the developed solution is the urban area noise mapping. The map can be updated completely within relatively short period of time, employing the PL-Grid Infrastructure. Operations are performed employing a dedicated noise prediction model, optimized for a computer cluster. In addition, predicted maps may be adjusted, using real noise level measurements. Animation: the fragment of the calculated dynamic noise map of the city of Gdansk, Poland
Ecology Automatic Phenology Observations Service (ApheS) 22 Automatic phenology observations infrastructure The ApheS service is based on the KIWI Remote Instrumentation Platform - a framework for building remote instrumentation systems for equipment and meteorological sensors Devices and sensors produce different output data. Most of the sensors produce numerical data, which is to be visualized by the PLGrid Plus Ecogrid Web Portal Example of photo taken by KIWI Eye observation tool
Ecology KIWI universal monitoring platform 23 Achievements: KNOW-HOW associated with the implementation of the observational sets for the WLIN project Observation eq uipment, hardware, software, materials, technologies, problems and their solutions Video: Prototype testing and the first results of an observation operating set Video: System run and its visual effects
Metallurgy Simulations of extrusion process in 3D deployed 25 Main Objectives: Optimization of the metallurgical process of profiles extrusion. Optimization includes: shape of foramera, channel position on a die, calibration stripes, extrusion velocity, ingot temperatures, tools. The proposed grid-based software simulates extrusion of thin profiles and rods of special alloys of magnesium, containing calcium supplements. These alloys are characterized by extremely low technological plasticity during metal forming. The FEM mathematical model developed.
Metallurgy Deployed domain specific services 26 SSRVE the service for generation a Statistically Representative Volume Element (called SSRVE) for the microstructures of materials. Representation of microstructure of two-phase metallic materials ModFemNet the service for welding simulations.
Chemistry Deployed domain specific service 29 InSilicoLab for Chemistry the service that aims to support the launch of complex computational quantum chemistry experiments in the PL-Grid Infrastructure. Experiments of this service make it easy to plan sequential computation schemes that require the preparation of series of data files, based on a common schema. Animation showing the service run
AstroGrid-PL core sevices Polish Virtual Observatory and Workflow Environment for Astronomy 30 Whole-community astronomical grid Main Polish astronomical institutes involved CAMK PAN (coordinator), CA UMK, OA UJ and a few others The Virtual Observatory (VObs) is an international astronomical community-based initiative aiming to provide transparent and distributed access to data available worldwide. Service goals: Setup National VObs Data Center integration of Polish data and join international efforts Workflow Environment Provide astronomers with workflow environment Kepler environment is being tested... Universal Fluid Dynamics code Piernik InSilicoLab for Astrophysics the service that aims to support the launch of complex astrophysical computational experiments in the PL-Grid Infrastructure.
Deployed domain specific services SynchroGrid 33 Elegant the service for those involved in the design and operation of Synchrotron. The service consists in: provision of the elegant (ELEctron Generation ANd Tracking) application in the parallel version on a cluster, configuring the Matlab software to read output files produced by this application in a Self Describing Data Sets (SDDS) format and to generate the final results in the form of drawings. Animation: simulation of the Synchrotron run Elegant is a fully 6D accelerator simulation program that now does much more than generation of particle distributions and tracking them
Energy Sector πesa - Platform for Integrated Energy System Analysis Service to be deployed in 2014 πesa narzędzie do budowy modeli systemów energetycznych TIMES, system modelowania jakości powietrza Polyphemus oraz model do oceny ich oddziaływania na środowisko i zdrowie ludzkie MAEH. OptiMINE analiza wariantów prowadzenia robót górniczych i wybór najlepszego z nich pod względem planowanego wydobycia węgla kamiennego w kopalni. Algorytm obliczeniowy opiera się na selekcji klonalnej i innych wybranych elementach sztucznych systemów immunologicznych. Wartości postępów w poszczególnych wyrobiskach i rozłożenie robót w czasie, których dotrzymanie zapewni osiągnięcie planowanego wydobycia. ModWELL Video: Atmosphere transport of air pollution Analiza funkcjonowania sektora wytwarzania energii elektrycznej w horyzoncie krótkoterminowym w wysokiej rozdzielczości. Z pomocą Wirtualnego Laboratorium GridSpace 2 34
Deployed domain specific services HEPGrid (High Energy Physics) 35 CVMFS the service that provides catalogs of software and data needed to reconstruct and analyse data in the HEP experiments. The service operates on a dedicated server installation on a readonly virtual file system CERNVM-FS, installed by the FUSE module in the local user space. With this service, there are immediately available all versions of the software and any modifications made at the central servers, at economical consumption of the local storage resources.
Future Development 36 New Computer Ecosystem Problems (to be) tackled Unified Portal Technologies for OpenScience Interactive Data processing Transparent access to data SSO Community Portal Community Portal Community Portal Accounting Unified Acess Portals Helpdesk Platform for HPC with Workflows Data-Farming Large-scale file systems Large-scale databases Cloud Services Computational Grids Computation and Data Processing Services PaaS for scientists MapReduce environment CYFRONET ICM PCSS TASK WCSS Geographically Distributed Infrastructure Support for strategic scientific projects (ESFRI, etc.)
Conclusions 37 Further developement needed, as identified currently, mainly on Domain Specific Grids Request from the users communities Capacity for organization of future development according to Expertise and experience Strong scientific potential of the users communities being represented by PL-Grid Consortium Wide international cooperation concerning the Consortium and individual Partners, good recognition worldwide Good managerial capacity Please visit our Web pages: http://www.plgrid.pl/en http://www.plgrid.pl Credits
Credits 38 ACC Cyfronet AGH Michał Turała Marian Bubak Krzysztof Zieliński Karol Krawentek Agnieszka Szymańska Maciej Twardy Teresa Ozga Angelika Zaleska-Walterbach Andrzej Oziębło Zofia Mosurska Marcin Radecki Renata Słota Tomasz Gubała Darin Nikolow Aleksandra Pałuk Patryk Lasoń Marek Magryś Łukasz Flis and many others domain experts. ICM Marek Niezgódka Piotr Bała Maciej Filocha PCSS Maciej Stroiński Norbert Meyer Krzysztof Kurowski Bartek Palak Tomasz Piontek Dawid Szejnfeld Paweł Wolniewicz WCSS Józef Janyszek Mateusz Tykierko Paweł Dziekoński Bartłomiej Balcerek TASK Rafał Tylman Mścislaw Nakonieczny Jarosław Rybicki