IT4Innovations national supercomputing center. Branislav Jansík

Transcription

1 IT4Innovations national supercomputing center Branislav Jansík

2 Anselm

3 Salomon

4 Data center infrastructure

5 Anselm and Salomon Anselm Intel Sandy Bridge E x8 cores 64GB RAM at 60 GB/s 209 nodes 23 GPGPU Nvidia K20 4 MIC Intel Xeon Phi Infiniband QDR (40 Gb/s) Fat tree TB storage 330 GB local storage 2x 512 GB, 16 core Fat node Salomon Intel Haswell E5-2680v3 2x12 cores 128GB RAM at 100 GB/s 1008 nodes 0 GPGPU Nvidia K MIC Intel Xeon Phi Infiniband FDR (56 Gb/s) 7D hypercube TB storage no local storage 1x 3328 GB, 112 core UV2000

6 Salomon vs Anselm Salomon Anselm

7 Intel Xeon Phi accelerator cards Xeon Phi 7120P x86 architecture, 1.2TF 864x Intel Xeon Phi 7120P 61(244) cores, 512 bit FMA 16 GB RAM

8 3.36TF per accelerated node 0.48TF 1.2TF 0.48TF Los Alamos Blue Mountain (2000) 3.07TF 1.2TF

9 UV 2000 SMP Specs 8 servers, dual socket 14x Intel Xeon E5-4627v2, 8-core 3.3GHz processors 14 NUMA nodes 112 cores in total 3328 GB of physical memory 1x NVIDIA GM200 (GeForce GTX TITAN X), 12GB RAM Access Name: uv1 Queue: qfat SMP NUMA Only AVX instructions. Binaries compiled for Salomon compute nodes do not run!

10 Available software Chemistry, bioinformatics CP2K, NWCHEM, MOLPRO, WIEN2K, VASP, BOWTIE2, BIOCONDUCTOR OMICS MASTER, FIJI Compilers INTEL, GNU, UPC, JAVA, (PGI), NVIDIA CUDA Engineering ADAMS, ANSYS, COMSOL, CODE_SATURNE, HYPERMESH, ENSIGHT, MATLAB, OCTAVE, PARAVIEW, VISIT, LSDYNA, MODFLOW, OPENFOAM, ELMER, NASTRAN, parallel R USE.OWN anything Libs MKL, FFTW, GPI2, GSL, HDF5, MPI, NETCDF, ADIOS, MAGMA, LIBMESH, PETSC, TRILINOS, SLEPC, TBB, IPP Tools ALLINEA DDT/MAP/PERF Reports, TOTALVIEW, VAMPIR, SCALASCA2, VALGRIND, SCOREP, VTUNE, IDB Virtualization KVM/QEMU, WINE, VDE2

11 Main services HPC resources Application and user support Training and documentation

12 Accessing HPC resources Grant competitions Utilization Grant competitions Utilization Epilogue Grant competitions Utilization Epilogue Epilogue

13 Accessing HPC resources Grant competitions Open Access 3x per year Directors discretion Utilization System access and accounts Code deployment Execution Free resources scavenging Epilogue Report and feedback Publication registration

14 Proposal content and evaluation Aims and objectives Methods and state-of-the-art Impact and outlooks Scientific readiness max 2 pages 1-5 Computational approach Parallelization and scalability Resources requested Computational readiness max 1 pages 1-5 Economic readiness 0-5 Socioeconomic impact Operational costs Amortization and renewal

15 Obtaining login credentials Authorization chain Collaborator 1 Allocation Committee Resource Allocation Authorization to utilize resources Authorization to utilize resources PI Collaborator 2 Authorization to utilize resources Collaborator n

16 Support

17 Documentation

18 Documentation

19 Self service

20 User support request tracker

21 Training

22 Training training events in past 12 months Diverse HPC domains are covered Top notch, expert speakers Highly recommended

23 Conclusions Computational resources available to general academic community are allocated in competition of scientific and technical quality Read the documentation, contact support, contact me! IT4Innovations SuperComputer Center is here to run the computer and assist you in using it

24 Open access at IT4I Allocation committee meets 3x per year Call for proposals core hours for period of 9 months Call will be closed after 40 working days Evaluation: Scientific, technical and economic assessment, external reviewers may be invited In 20 working days, evaluation will be available 10 working days for rebuttal Results announced in 5 working days Utilization 9 months Summary report expected within 12 months