COSC 6374 Parallel Computation. Scientific Data Libraries. Edgar Gabriel Fall Motivation
|
|
- Brianna Stevenson
- 6 years ago
- Views:
Transcription
1 COSC 6374 Parallel Computation Scientific Data Libraries Edgar Gabriel Fall 2013 Motivation MPI I/O is good It knows about data types (=> data conversion) It can optimize various access patterns in applications MPI I/O is bad It does not store any information about the data type A file written as MPI_INT can be read as MPI_DOUBLE in another application No information is stored about data layout, e.g. two-dimensional data set, extent of each dimension etc. 1
2 Scientific data libraries Handle data on a higher level Provide additional information (Metadata) Size and type of of data structure Data format Name Units Two widely used libraries available NetCDF HDF-5 HDF-5 Hierarchical Data Format (HDF) developed since 1988 at NCSA (University of Illinois) Has gone through a long history of changes, the recent version HDF-5 available since 1999 HDF-5 supports Very large files Parallel I/O interface Fortran, C, Java bindings 2
3 HDF-5 dataset Multi-dimensional array of basic data elements A dataset consists of Header + data Header consists of Name Datatype : basic (e.g. HDF_NATIVE_FLOAT) or compound dataypes Dataspace: defines size and shape of a multidimensional array. Dimensions can be fixed or unlimited. Storage layout: defines how multidimensional arrays are stored in file. Can be contiguous or chunked. Example of an HDF-5 file HDF5 tempseries.h5 { GROUP / { GROUP tempseries { DATASET height { DATATYPE { H5_STD_I32BE } DATASPACE ( ARRAY (4) (4) } DATA { 0, 50, 100, 150 } ATTRIBUTES units { DATATYPE { undefined string } DATASPACE { ARRAY (0) (0) } DATA { unable to print } } } DATASET temperature { DATATYPE { H5T_IEEE_F32BE } DATASPACE{ ARRAY( 3,8,4 ) (H5S_UNLIMITED, 8, 4) } DATA { } 3
4 Storage layout: contiguous vs. chunked contiguous chunked Advantages and disadvantages of chunking Accessing rows and columns require the same number of accesses Data can be extended into all dimensions Efficient storage of sparse arrays Can improve caching HDF-5 API HDF-5 naming convention All API functions start with an H5 The next character identifies category of functions H5F: functions handling files H5G: functions handling groups H5D: functions handling datasets H5S: functions handling dataspaces H5A: functions handling attributes A HDF-5 group is a collection of data sets Comparable to a directory in a UNIX-like file system 4
5 Writing a sequential HDF-5 file 1. Create the file 2. Create a group (opt.) 3. Define a dataspace 4. Define datatype 5. Create dataset 6. Add attributes h5file = H5Fcreate( ) group = H5Gcreate (h5file, ) tspace = H5Screate_simple(ndims, dims, maxdims ); ttype = H5T_IEEE_F32BE; tset = H5Dcreate (group, testset, ttype, tspace, ); tattr = H5Acreate (tset, units, H5T_C_S1, ) H5Awrite (tattr, H5T_C_S1, meter ); H5Dwrite(tset,H5T_IEEE_F32BE,,data); 7. Write data 8. Close all objects Reading an HDF-5 file structure of the file known 1. Open the file 2. Open the group 3. Open dataset in the group 4. Look up dimensions 5. Read data 6. Read attributes 7. Read comments 8. Close all objects h5file = H5Fopen( ) group = H5Gopen(h5file, tempseries ) tset = H5Dopen(group, temperature ); tspace = H5Dget_space( tset ); H5Sget_simple_extent_dims (tspace, dims, ); H5Dread(tset,H5T_IEEE_F32BE, ttype, tspace,, buffer); tattr = H5Aopen_name(tset, units ); attrtype = H5Aget_type ( tattr ); H5Aread(tattr,attrtype,attr); 5
6 Compound Datatypes Abstraction for user structures Has a fixed size Each member has its own name, datatype, reference, and byte offset h5type = H5Tcreate( H5T_class class, size_t size); H5Tinsert ( h5type, const char *name, off_t offset, hid_t field_id); Hyperslab A hyperslab is a portion of a dataset H5Sselect_hyperslab (hid_t space_id, H5S_seloper_t operator, const hssize_t *start, const hsize_t *stride, const hsize_t *count, const hsize_t *block); Operator: H5S_SELECT_SET, H5S_SELECT_OR Start: array determining the starting coordinates of the hyperslab Stride: array indicating which elements along a dimension are to be selected Count: array determining how many points to use in each dimension Block: array determining the size of the element block by the datatype 6
7 Example using hyperslabs /* Define hyperslab in the dataset. */ offset[0] = 1; count[0] = NX_SUB; offset[1] = 2; count[1] = NY_SUB; status = H5Sselect_hyperslab (dataspace, H5S_SELECT_SET, offset, NULL, count, NULL); /*Read data from hyperslab in file into hyperslab in memory */ status = H5Dread (dataset, H5T_NATIVE_INT, memspace, dataspace, H5P_DEFAULT, data_out); offset[0] count[0] offset[1] count[1] Examples taken from HDF-5 webpage Process 0 More complex example using hyperslabs Memory offset[0] offset[1] on rank 0 1 File count[0] = 1; count[1] = dimsmem[1]; block[0] = dimsfile[0]; block[1] = 1; offset[0] = 0; offset[1] = mpi_rank; stride[0] = 1; stride[1] = 2; Process 1 dimsmem[0] dimsmem[1] block[0] For dimension x: you generate count[x] entries of block[x] elements starting from offset[x]. The distance between each element is stride[x]. Examples taken from HDF-5 webpage 7
8 Parallel I/O with HDF-5 Relies on MPI I/O Program has to use special properties (hints) indicating to use parallel I/O during File creation and file open Data access Properties are set through H5P functions /* example for using file properties */ fileprops = H5Pcreate (H5P_FILE_ACCESS); H5Pset_fapl_mpio ( fileprops, MPI_COMM_WORLD, MPI_INFO_NULL); h5file = H5Fcreate ( tempseries.h5,, fileprops); Parallel data access in HDF-5 Application has to define a set of interleaved file data spaces on the processes that will access the file Similar technique like setting the file-view in MPI I/O Usually based on defining hyperslabs Data transfer properties have to be set /* example for using file properties */ fileprops = H5Pcreate (H5P_DATASET_XFER); H5Pset_dxpl_mpio ( fileprops, H5FD_MPIO_COLLECTIVE); H5Dwrite ( tset,, fileprops); 8
Hierarchical Data Format 5:
Hierarchical Data Format 5: Giusy Muscianisi g.muscianisi@cineca.it SuperComputing Applications and Innovation Department May 17th, 2013 Outline What is HDF5? Overview to HDF5 Data Model and File Structure
More informationParallel HDF5 (PHDF5)
Parallel HDF5 (PHDF5) Giusy Muscianisi g.muscianisi@cineca.it SuperComputing Applications and Innovation Department May 17th, 2013 Outline Overview of Parallel HDF5 design Programming model for Creating
More informationParallel I/O and Portable Data Formats HDF5
Parallel I/O and Portable Data Formats HDF5 Sebastian Lührs s.luehrs@fz-juelich.de Jülich Supercomputing Centre Forschungszentrum Jülich GmbH Jülich, March 13th, 2018 Outline Introduction Structure of
More informationIntroduction to HDF5
The HDF Group Introduction to HDF5 Quincey Koziol Director of Core Software & HPC The HDF Group October 15, 2014 Blue Waters Advanced User Workshop 1 Why HDF5? Have you ever asked yourself: How will I
More informationHDF5: An Introduction. Adam Carter EPCC, The University of Edinburgh
HDF5: An Introduction Adam Carter EPCC, The University of Edinburgh What is HDF5? Hierarchical Data Format (version 5) From www.hdfgroup.org: HDF5 is a unique technology suite that makes possible the management
More informationParallel I/O and Portable Data Formats
Parallel I/O and Portable Data Formats Sebastian Lührs s.luehrs@fz-juelich.de Jülich Supercomputing Centre Forschungszentrum Jülich GmbH Reykjavík, August 25 th, 2017 Overview I/O can be the main bottleneck
More informationHDF5: theory & practice
HDF5: theory & practice Giorgio Amati SCAI Dept. 15/16 May 2014 Agenda HDF5: main issues Using the API (serial) Using the API (parallel) Tools Some comments PHDF5 Initial Target Support for MPI programming
More informationRFC: HDF5 Virtual Dataset
RFC: HDF5 Virtual Dataset Quincey Koziol (koziol@hdfgroup.org) Elena Pourmal (epourmal@hdfgroup.org) Neil Fortner (nfortne2@hdfgroup.org) This document introduces Virtual Datasets (VDS) for HDF5 and summarizes
More informationUsing HDF5 for Scientific Data Analysis. NERSC Visualization Group
Using HDF5 for Scientific Data Analysis NERSC Visualization Group Before We Get Started Glossary of Terms Data - The raw information expressed in numerical form Metadata - Ancillary information about your
More informationParallel I/O CPS343. Spring Parallel and High Performance Computing. CPS343 (Parallel and HPC) Parallel I/O Spring / 22
Parallel I/O CPS343 Parallel and High Performance Computing Spring 2018 CPS343 (Parallel and HPC) Parallel I/O Spring 2018 1 / 22 Outline 1 Overview of parallel I/O I/O strategies 2 MPI I/O 3 Parallel
More informationHDF5 User s Guide. HDF5 Release November
HDF5 User s Guide HDF5 Release 1.8.8 November 2011 http://www.hdfgroup.org Copyright Notice and License Terms for HDF5 (Hierarchical Data Format 5) Software Library and Utilities HDF5 (Hierarchical Data
More informationPackage rhdf5. April 5, 2014
Package rhdf5 April 5, 2014 Type Package Title HDF5 interface to R Version 2.6.0 Author, Gregoire Pau Maintainer This R/Bioconductor package provides an interface between HDF5 and
More informationIntroduction to HDF5
Introduction to parallel HDF Maison de la Simulation Saclay, 0-0 March 201, Parallel filesystems and parallel IO libraries PATC@MdS Evaluation form Please do not forget to fill the evaluation form at https://events.prace-ri.eu/event/30/evaluation/evaluate
More informationIntroduction to serial HDF5
Introduction to serial HDF Matthieu Haefele Saclay, - March 201, Parallel filesystems and parallel IO libraries PATC@MdS Matthieu Haefele Training outline Day 1: AM: Serial HDF (M. Haefele) PM: Parallel
More informationNtuple: Tabular Data in HDF5 with C++ Chris Green and Marc Paterno HDF5 Webinar,
Ntuple: Tabular Data in HDF5 with C++ Chris Green and Marc Paterno HDF5 Webinar, 2019-01-24 Origin and motivation Particle physics analysis often involves the creation of Ntuples, tables of (usually complicated)
More informationThe HDF Group. Parallel HDF5. Quincey Koziol Director of Core Software & HPC The HDF Group.
The HDF Group Parallel HDF5 Quincey Koziol Director of Core Software & HPC The HDF Group Parallel HDF5 Success Story Recent success story Trillion particle simulation on hopper @ NERSC 120,000 cores 30TB
More informationThe HDF Group. Parallel HDF5. Extreme Scale Computing Argonne.
The HDF Group Parallel HDF5 Advantage of Parallel HDF5 Recent success story Trillion particle simulation on hopper @ NERSC 120,000 cores 30TB file 23GB/sec average speed with 35GB/sec peaks (out of 40GB/sec
More informationCaching and Buffering in HDF5
Caching and Buffering in HDF5 September 9, 2008 SPEEDUP Workshop - HDF5 Tutorial 1 Software stack Life cycle: What happens to data when it is transferred from application buffer to HDF5 file and from HDF5
More informationIntroduction to I/O at CHPC
CENTER FOR HIGH PERFORMANCE COMPUTING Introduction to I/O at CHPC Martin Čuma, m.cumautah.edu Center for High Performance Computing Fall 2015 Outline Types of storage available at CHPC Types of file I/O
More informationDRAFT. HDF5 Data Flow Pipeline for H5Dread. 1 Introduction. 2 Examples
This document describes the HDF5 library s data movement and processing activities when H5Dread is called for a dataset with chunked storage. The document provides an overview of how memory management,
More informationWhat NetCDF users should know about HDF5?
What NetCDF users should know about HDF5? Elena Pourmal The HDF Group July 20, 2007 7/23/07 1 Outline The HDF Group and HDF software HDF5 Data Model Using HDF5 tools to work with NetCDF-4 programs files
More informationPost-processing issue, introduction to HDF5
Post-processing issue Introduction to HDF5 Matthieu Haefele High Level Support Team Max-Planck-Institut für Plasmaphysik, München, Germany Autrans, 26-30 Septembre 2011, École d été Masse de données :
More informationHDF- A Suitable Scientific Data Format for Satellite Data Products
HDF- A Suitable Scientific Data Format for Satellite Data Products Sk. Sazid Mahammad, Debajyoti Dhar and R. Ramakrishnan Data Products Software Division Space Applications Centre, ISRO, Ahmedabad 380
More informationNew Features in HDF5. Why new features? September 9, 2008 SPEEDUP Workshop - HDF5 Tutorial
New Features in HDF5 September 9, 2008 SPEEDUP Workshop - HDF5 Tutorial 1 Why new features? September 9, 2008 SPEEDUP Workshop - HDF5 Tutorial 2 1 Why new features? HDF5 1.8.0 was released in February
More informationh5perf_serial, a Serial File System Benchmarking Tool
h5perf_serial, a Serial File System Benchmarking Tool The HDF Group April, 2009 HDF5 users have reported the need to perform serial benchmarking on systems without an MPI environment. The parallel benchmarking
More informationIntroduction to HDF5
Introduction to HDF5 Dr. Shelley L. Knuth Research Computing, CU-Boulder December 11, 2014 h/p://researchcompu7ng.github.io/meetup_fall_2014/ Download data used today from: h/p://neondataskills.org/hdf5/exploring-
More informationI/O in scientific applications
COSC 4397 Parallel I/O (II) Access patterns Spring 2010 I/O in scientific applications Different classes of I/O operations Required I/O: reading input data and writing final results Checkpointing: data
More informationIntroduction to I/O at CHPC
CENTER FOR HIGH PERFORMANCE COMPUTING Introduction to I/O at CHPC Martin Čuma, m.cuma@utah.edu Center for High Performance Computing Fall 2018 Outline Types of storage available at CHPC Types of file I/O
More informationJialin Liu, Evan Racah, Quincey Koziol, Richard Shane Canon, Alex Gittens, Lisa Gerhardt, Suren Byna, Mike F. Ringenburg, Prabhat
H5Spark H5Spark: Bridging the I/O Gap between Spark and Scien9fic Data Formats on HPC Systems Jialin Liu, Evan Racah, Quincey Koziol, Richard Shane Canon, Alex Gittens, Lisa Gerhardt, Suren Byna, Mike
More informationParallel I/O: Not Your Job. Rob Latham and Rob Ross Mathematics and Computer Science Division Argonne National Laboratory {robl,
Parallel I/O: Not Your Job Rob Latham and Rob Ross Mathematics and Computer Science Division Argonne National Laboratory {robl, rross}@mcs.anl.gov Computational Science! Use of computer simulation as a
More informationCOSC 6374 Parallel Computation. Introduction to MPI V Derived Data Types. Edgar Gabriel Fall Derived Datatypes
COSC 6374 Parallel Computation Introduction to MPI V Derived Data Types Edgar Gabriel Fall 2013 Derived Datatypes Basic idea: describe memory layout of user data structures e.g. a structure in C typedef
More informationImplementing HDF5 in MATLAB
Implementing HDF5 in MATLAB Jeff Mather & Alec Rogers The MathWorks, Inc. 2006 The MathWorks, Inc. 29 November 2006 HDF4 1-1 mapping of C API first. (1998) Customer requests for high-level functions. HDFREAD,
More informationAdapting Software to NetCDF's Enhanced Data Model
Adapting Software to NetCDF's Enhanced Data Model Russ Rew UCAR Unidata EGU, May 2010 Overview Background What is netcdf? What is the netcdf classic data model? What is the netcdf enhanced data model?
More informationHDF5 I/O Performance. HDF and HDF-EOS Workshop VI December 5, 2002
HDF5 I/O Performance HDF and HDF-EOS Workshop VI December 5, 2002 1 Goal of this talk Give an overview of the HDF5 Library tuning knobs for sequential and parallel performance 2 Challenging task HDF5 Library
More informationParallel I/O Performance Study and Optimizations with HDF5, A Scientific Data Package
Parallel I/O Performance Study and Optimizations with HDF5, A Scientific Data Package MuQun Yang, Christian Chilan, Albert Cheng, Quincey Koziol, Mike Folk, Leon Arber The HDF Group Champaign, IL 61820
More informationParallel I/O from a User s Perspective
Parallel I/O from a User s Perspective HPC Advisory Council Stanford University, Dec. 6, 2011 Katie Antypas Group Leader, NERSC User Services NERSC is DOE in HPC Production Computing Facility NERSC computing
More informationrhdf5 - HDF5 interface for R
Bernd Fischer October 30, 2017 Contents 1 Introduction 1 2 Installation of the HDF5 package 2 3 High level R -HDF5 functions 2 31 Creating an HDF5 file and group hierarchy 2 32 Writing and reading objects
More informationCOSC 6374 Parallel Computation. Derived Data Types in MPI. Edgar Gabriel. Spring Derived Datatypes
COSC 6374 Parallel Computation Derived Data Types in MPI Spring 2008 Derived Datatypes Basic idea: interface to describe memory layout of user data structures e.g. a structure in C typedef struct { char
More informationCOSC 6374 Parallel Computation. Remote Direct Memory Acces
COSC 6374 Parallel Computation Remote Direct Memory Acces Edgar Gabriel Fall 2013 Communication Models A P0 receive send B P1 Message Passing Model: Two-sided communication A P0 put B P1 Remote Memory
More informationIntroduction to NetCDF
Introduction to NetCDF NetCDF is a set of software libraries and machine-independent data formats that support the creation, access, and sharing of array-oriented scientific data. First released in 1989.
More informationNetCDF and Scientific Data Durability. Russ Rew, UCAR Unidata ESIP Federation Summer Meeting
NetCDF and Scientific Data Durability Russ Rew, UCAR Unidata ESIP Federation Summer Meeting 2009-07-08 For preserving data, is format obsolescence a non-issue? Why do formats (and their access software)
More informationParallel I/O and Portable Data Formats PnetCDF and NetCDF 4
Parallel I/O and Portable Data Formats PnetDF and NetDF 4 Sebastian Lührs s.luehrs@fz-juelich.de Jülich Supercomputing entre Forschungszentrum Jülich GmbH Jülich, March 13 th, 2017 Outline Introduction
More informationParallel NetCDF. Rob Latham Mathematics and Computer Science Division Argonne National Laboratory
Parallel NetCDF Rob Latham Mathematics and Computer Science Division Argonne National Laboratory robl@mcs.anl.gov I/O for Computational Science Application Application Parallel File System I/O Hardware
More informationNetCDF-4: : Software Implementing an Enhanced Data Model for the Geosciences
NetCDF-4: : Software Implementing an Enhanced Data Model for the Geosciences Russ Rew, Ed Hartnett, and John Caron UCAR Unidata Program, Boulder 2006-01-31 Acknowledgments This work was supported by the
More informationHDF5 File Space Management. 1. Introduction
HDF5 File Space Management 1. Introduction The space within an HDF5 file is called its file space. When a user first creates an HDF5 file, the HDF5 library immediately allocates space to store information
More information7C.2 EXPERIENCE WITH AN ENHANCED NETCDF DATA MODEL AND INTERFACE FOR SCIENTIFIC DATA ACCESS. Edward Hartnett*, and R. K. Rew UCAR, Boulder, CO
7C.2 EXPERIENCE WITH AN ENHANCED NETCDF DATA MODEL AND INTERFACE FOR SCIENTIFIC DATA ACCESS Edward Hartnett*, and R. K. Rew UCAR, Boulder, CO 1 INTRODUCTION TO NETCDF AND THE NETCDF-4 PROJECT The purpose
More informationMapping HDF4 Objects to HDF5 Objects
Mapping HDF4 bjects to HDF5 bjects Mike Folk, Robert E. McGrath, Kent Yang National Center for Supercomputing Applications, University of Illinois February, 2000 Revised: ctober, 2000 Note to reader: We
More informationAn Evolutionary Path to Object Storage Access
An Evolutionary Path to Object Storage Access David Goodell +, Seong Jo (Shawn) Kim*, Robert Latham +, Mahmut Kandemir*, and Robert Ross + *Pennsylvania State University + Argonne National Laboratory Outline
More informationLecture 33: More on MPI I/O. William Gropp
Lecture 33: More on MPI I/O William Gropp www.cs.illinois.edu/~wgropp Today s Topics High level parallel I/O libraries Options for efficient I/O Example of I/O for a distributed array Understanding why
More informationThe netcdf- 4 data model and format. Russ Rew, UCAR Unidata NetCDF Workshop 25 October 2012
The netcdf- 4 data model and format Russ Rew, UCAR Unidata NetCDF Workshop 25 October 2012 NetCDF data models, formats, APIs Data models for scienbfic data and metadata - classic: simplest model - - dimensions,
More informationCOSC 6397 Big Data Analytics. Distributed File Systems (II) Edgar Gabriel Fall HDFS Basics
COSC 6397 Big Data Analytics Distributed File Systems (II) Edgar Gabriel Fall 2018 HDFS Basics An open-source implementation of Google File System Assume that node failure rate is high Assumes a small
More informationObject-Oriented Programming for Scientific Computing
Object-Oriented Programming for Scientific Computing Traits and Policies Ole Klein Interdisciplinary Center for Scientific Computing Heidelberg University ole.klein@iwr.uni-heidelberg.de 11. Juli 2017
More informationHDF5 C++ User s Notes
HDF5 C++ User s Notes This User s Note provides an overview of the structure, the availability, and the limitations of the C++ API of HDF5. It lists the classes and member functions included in the API
More informationIntroduction to Parallel I/O
Introduction to Parallel I/O Bilel Hadri bhadri@utk.edu NICS Scientific Computing Group OLCF/NICS Fall Training October 19 th, 2011 Outline Introduction to I/O Path from Application to File System Common
More informationCOSC 6374 Parallel Computation. Remote Direct Memory Access
COSC 6374 Parallel Computation Remote Direct Memory Access Edgar Gabriel Fall 2015 Communication Models A P0 receive send B P1 Message Passing Model A B Shared Memory Model P0 A=B P1 A P0 put B P1 Remote
More informationAn Overview of the HDF5 Technology Suite and its Applications
An Overview of the HDF5 Technology Suite and its Applications Mike Folk, Gerd Heber, Quincey Koziol, Elena Pourmal, Dana Robinson The HDF Group {mfolk,gheber,koziol,epourmal,derobins}@hdfgroup.org Outline
More informationWhat is a file system
COSC 6397 Big Data Analytics Distributed File Systems Edgar Gabriel Spring 2017 What is a file system A clearly defined method that the OS uses to store, catalog and retrieve files Manage the bits that
More informationReusing this material
Derived Datatypes Reusing this material This work is licensed under a Creative Commons Attribution- NonCommercial-ShareAlike 4.0 International License. http://creativecommons.org/licenses/by-nc-sa/4.0/deed.en_us
More informationMapping HDF4 Objects to HDF5 Objects Version 3
Mapping HDF4 Objects to HDF5 Objects Version 3 Mike Folk, Robert E. McGrath, Kent Yang National Center for Supercomputing Applications University of Illinois, Urbana-Champaign February, 2000 Revised: October,
More informationMapping HDF4 Objects to HDF5 Objects Version 4
Mapping HDF4 bjects to HDF5 bjects Version 4 Mike Folk 1, Robert E. McGrath 2, Kent Yang 1 Revised: ctober, 2000; July, 2002, August, 2003, September 2017 1 The HDF Group 2 National Center for Supercomputing
More informationAPI and Usage of libhio on XC-40 Systems
API and Usage of libhio on XC-40 Systems May 24, 2018 Nathan Hjelm Cray Users Group May 24, 2018 Los Alamos National Laboratory LA-UR-18-24513 5/24/2018 1 Outline Background HIO Design HIO API HIO Configuration
More informationHigh-Performance Techniques for Parallel I/O
1 High-Performance Techniques for Parallel I/O Avery Ching Northwestern University Kenin Coloma Northwestern University Jianwei Li Northwestern University Alok Choudhary Northwestern University Wei-keng
More informationThe HDF Group Q5 Demo
The HDF Group The HDF Group Q5 Demo 5.6 HDF5 Transaction API 5.7 Full HDF5 Dynamic Data Structure NOTICE: THIS MANUSCRIPT HAS BEEN AUTHORED BY INTEL UNDER ITS SUBCONTRACT WITH LAWRENCE LIVERMORE NATIONAL
More informationRFC: HDF5 File Space Management: Paged Aggregation
RFC: HDF5 File Space Management: Paged Aggregation Vailin Choi Quincey Koziol John Mainzer The current HDF5 file space allocation accumulates small pieces of metadata and raw data in aggregator blocks.
More informationMPI: A Message-Passing Interface Standard
MPI: A Message-Passing Interface Standard Version 2.1 Message Passing Interface Forum June 23, 2008 Contents Acknowledgments xvl1 1 Introduction to MPI 1 1.1 Overview and Goals 1 1.2 Background of MPI-1.0
More informationA FRAMEWORK ARCHITECTURE FOR SHARED FILE POINTER OPERATIONS IN OPEN MPI
A FRAMEWORK ARCHITECTURE FOR SHARED FILE POINTER OPERATIONS IN OPEN MPI A Thesis Presented to the Faculty of the Department of Computer Science University of Houston In Partial Fulfillment of the Requirements
More informationA Plugin for HDF5 using PLFS for Improved I/O Performance and Semantic Analysis
2012 SC Companion: High Performance Computing, Networking Storage and Analysis A for HDF5 using PLFS for Improved I/O Performance and Semantic Analysis Kshitij Mehta, John Bent, Aaron Torres, Gary Grider,
More informationDESY IT Seminar HDF5, Nexus, and what it is all about
DESY IT Seminar HDF5, Nexus, and what it is all about Eugen Wintersberger HDF5 and Nexus DESY IT, 27.05.2013 Why should we care about Nexus and HDF5? Current state: Data is stored either as ASCII file
More informationECSS Project: Prof. Bodony: CFD, Aeroacoustics
ECSS Project: Prof. Bodony: CFD, Aeroacoustics Robert McLay The Texas Advanced Computing Center June 19, 2012 ECSS Project: Bodony Aeroacoustics Program Program s name is RocfloCM It is mixture of Fortran
More informationDistributed Memory Parallel Programming
COSC Big Data Analytics Parallel Programming using MPI Edgar Gabriel Spring 201 Distributed Memory Parallel Programming Vast majority of clusters are homogeneous Necessitated by the complexity of maintaining
More informationMilestone 8.1: HDF5 Index Demonstration
The HDF Group Milestone 8.1: HDF5 Index Demonstration Ruth Aydt, Mohamad Chaarawi, Quincey Koziol, Aleksandar Jelenak, Jerome Soumagne 06/30/2014 NOTICE: THIS MANUSCRIPT HAS BEEN AUTHORED BY THE HDF GROUP
More informationUsing Chunked Extendible Array for Physical Storage of Scientific Datasets
Using Chunked Extendible Array for Physical Storage of Scientific Datasets E.J Otoo, G. Nimako and D. Ohene-Kwofie School of Computer Science The University of the Witwatersrand Johannesburg, South Africa
More informationCS4961 Parallel Programming. Lecture 16: Introduction to Message Passing 11/3/11. Administrative. Mary Hall November 3, 2011.
CS4961 Parallel Programming Lecture 16: Introduction to Message Passing Administrative Next programming assignment due on Monday, Nov. 7 at midnight Need to define teams and have initial conversation with
More informationIntroduction to MPI Programming Part 2
Introduction to MPI Programming Part 2 Outline Collective communication Derived data types Collective Communication Collective communications involves all processes in a communicator One to all, all to
More informationFile Layout and Directories
COS 318: Operating Systems File Layout and Directories Jaswinder Pal Singh Computer Science Department Princeton University (http://www.cs.princeton.edu/courses/cos318/) Topics File system structure Disk
More informationCOSC 6374 Parallel Computation. Message Passing Interface (MPI ) I Introduction. Distributed memory machines
Network card Network card 1 COSC 6374 Parallel Computation Message Passing Interface (MPI ) I Introduction Edgar Gabriel Fall 015 Distributed memory machines Each compute node represents an independent
More informationFile System Implementation
File System Implementation Jinkyu Jeong (jinkyu@skku.edu) Computer Systems Laboratory Sungkyunkwan University http://csl.skku.edu SSE3044: Operating Systems, Fall 2016, Jinkyu Jeong (jinkyu@skku.edu) Implementing
More informationChapter 11. Parallel I/O. Rajeev Thakur and William Gropp
Chapter 11 Parallel I/O Rajeev Thakur and William Gropp Many parallel applications need to access large amounts of data. In such applications, the I/O performance can play a significant role in the overall
More informationLaplace Exercise Solution Review
Laplace Exercise Solution Review John Urbanic Parallel Computing Scientist Pittsburgh Supercomputing Center Copyright 2017 Finished? If you have finished, we can review a few principles that you have inevitably
More informationThe Message Passing Interface (MPI) TMA4280 Introduction to Supercomputing
The Message Passing Interface (MPI) TMA4280 Introduction to Supercomputing NTNU, IMF January 16. 2017 1 Parallelism Decompose the execution into several tasks according to the work to be done: Function/Task
More informationFile System Implementation. Sunu Wibirama
File System Implementation Sunu Wibirama File-System Structure Outline File-System Implementation Directory Implementation Allocation Methods Free-Space Management Discussion File System Structure File
More informationA Parallel API for Creating and Reading NetCDF Files
A Parallel API for Creating and Reading NetCDF Files January 4, 2015 Abstract Scientists recognize the importance of portable and efficient mechanisms for storing datasets created and used by their applications.
More informationHDF Product Designer: A tool for building HDF5 containers with granule metadata
The HDF Group HDF Product Designer: A tool for building HDF5 containers with granule metadata Lindsay Powers Aleksandar Jelenak, Joe Lee, Ted Habermann The HDF Group Data Producer s Conundrum 2 HDF Features
More informationCOSC 6374 Parallel Computation. Parallel I/O (I) I/O basics. Concept of a clusters
COSC 6374 Parallel I/O (I) I/O basics Fall 2010 Concept of a clusters Processor 1 local disks Compute node message passing network administrative network Memory Processor 2 Network card 1 Network card
More informationHierarchical Data Format query language (HDFql)
Hierarchical Data Format query language (HDFql) Reference Manual Version 1.5.0 December 2017 Copyright (C) 2016-2017 This document is part of the Hierarchical Data Format query language (HDFql). For more
More informationWelcome! Virtual tutorial starts at 15:00 BST
Welcome! Virtual tutorial starts at 15:00 BST Parallel IO and the ARCHER Filesystem ARCHER Virtual Tutorial, Wed 8 th Oct 2014 David Henty Reusing this material This work is licensed
More informationCOSC 6397 Big Data Analytics. Distributed File Systems (II) Edgar Gabriel Spring HDFS Basics
COSC 6397 Big Data Analytics Distributed File Systems (II) Edgar Gabriel Spring 2017 HDFS Basics An open-source implementation of Google File System Assume that node failure rate is high Assumes a small
More informationTechnical Specification on further interoperability with C
Technical Specification on further interoperability with C John Reid, ISO Fortran Convener Fortran 2003 (or 2008) provides for interoperability of procedures with nonoptional arguments that are scalars,
More informationMessage-Passing and MPI Programming
Message-Passing and MPI Programming 2.1 Transfer Procedures Datatypes and Collectives N.M. Maclaren Computing Service nmm1@cam.ac.uk ext. 34761 July 2010 These are the procedures that actually transfer
More informationAdvanced Parallel Programming
Advanced Parallel Programming Derived Datatypes Dr Daniel Holmes Applications Consultant dholmes@epcc.ed.ac.uk Overview Lecture will cover derived datatypes memory layouts vector datatypes floating vs
More informationAdvanced Parallel Programming
Advanced Parallel Programming Derived Datatypes Dr David Henty HPC Training and Support Manager d.henty@epcc.ed.ac.uk +44 131 650 5960 16/01/2014 MPI-IO 2: Derived Datatypes 2 Overview Lecture will cover
More informationPreview. COSC350 System Software, Fall
Preview File System File Name, File Structure, File Types, File Access, File Attributes, File Operation Directories Directory Operations File System Layout Implementing File Contiguous Allocation Linked
More informationMPI Parallel I/O. Chieh-Sen (Jason) Huang. Department of Applied Mathematics. National Sun Yat-sen University
MPI Parallel I/O Chieh-Sen (Jason) Huang Department of Applied Mathematics National Sun Yat-sen University Materials are taken from the book, Using MPI-2: Advanced Features of the Message-Passing Interface
More informationHigh Performance Computing Course Notes Message Passing Programming III
High Performance Computing Course Notes 2008-2009 2009 Message Passing Programming III Communication modes Synchronous mode The communication is considered complete when the sender receives the acknowledgement
More informationIntroduction to MPI-2 (Message-Passing Interface)
Introduction to MPI-2 (Message-Passing Interface) What are the major new features in MPI-2? Parallel I/O Remote Memory Operations Dynamic Process Management Support for Multithreading Parallel I/O Includes
More informationFile System Implementation. Jin-Soo Kim Computer Systems Laboratory Sungkyunkwan University
File System Implementation Jin-Soo Kim (jinsookim@skku.edu) Computer Systems Laboratory Sungkyunkwan University http://csl.skku.edu Implementing a File System On-disk structures How does file system represent
More informationChapter 4. Message-passing Model
Chapter 4 Message-Passing Programming Message-passing Model 2 1 Characteristics of Processes Number is specified at start-up time Remains constant throughout the execution of program All execute same program
More informationParallel I/O Libraries and Techniques
Parallel I/O Libraries and Techniques Mark Howison User Services & Support I/O for scientifc data I/O is commonly used by scientific applications to: Store numerical output from simulations Load initial
More informationPractical Scientific Computing: Performanceoptimized
Practical Scientific Computing: Performanceoptimized Programming Advanced MPI Programming December 13, 2006 Dr. Ralf-Peter Mundani Department of Computer Science Chair V Technische Universität München,
More informationC for Engineers and Scientists: An Interpretive Approach. Chapter 10: Arrays
Chapter 10: Arrays 10.1 Declaration of Arrays 10.2 How arrays are stored in memory One dimensional (1D) array type name[expr]; type is a data type, e.g. int, char, float name is a valid identifier (cannot
More information