-

This site is deprecated and will be decommissioned shortly. For current information regarding HPC visit our new site: hpc.njit.edu

Difference between pages "ModulesOnLochness" and "NJHPCFEDAuriStorEval"

From NJIT-ARCS HPC Wiki
(Difference between pages)
Jump to: navigation, search
(Importing text file)
 
(Importing text file)
 
Line 1: Line 1:
==Lmod==
 
  
Lmod is used to to manage modules on Lochness and Stheno. The software stack on Lochness and Stheno is new. Submit scripts will need to be adjusted to account for the new lmod system.  
+
<strong>Objective</strong>: RU, MSU, NJIT technical staffs judge the value that AuriStorFS
 +
would likely bring to the NJHPCFED, and later to the Eastern Regional Research
 +
Platform (ERRP).
  
For more information on Lmod check the [https://lmod.readthedocs.io/en/latest/010_user.html user manual].
+
If that value is considered sufficient, proceed with budgeting the purchase of AuriStorFS.
  
===module available===
+
<ol>
 +
<li>Go over in detail the items listed below</li>
 +
<ul>
 +
  <li>Designed for wide area network </li>
 +
  <li>Single global namespace for all clients. All clients see exactly the same path and ACLs.</li>
 +
  <li>Clients for all platforms - Linux, MacOS X, Windows, others</li>
 +
  <li>Extremely efficient administration and applications distribution
 +
  <ul>
 +
  <li>Administration by any authorized Kerberos principal from any AFS client</li>
 +
  </ul>
 +
  <li>Long history of working well and reliably at many institutions and international corporations</li>
 +
  <li>No administrator intervention in making mount points available to all
 +
    clients, other than creating the mount point - a single command done from any AFS client</li>
 +
  <li>Volume : basic storage unit. Volumes work very well as specific-purpose containers, e.g., software, datasets,
 +
  web content, documentation)</li>
 +
  <li>Read-only replication of volumes</li>
 +
  <li>Scalability - number of clients, volumes, users are readily accommodated</li>
 +
  <li>Fine-grained ACLs [https://computing.cs.cmu.edu/help-support/afs-acls.html CMU doc]</li>
 +
  <ul>
 +
<li>Directory</li>
 +
<ul>
 +
<li><strong>l</strong> (lookup) List the contents of a directory;  does not allow the reading of files</li>
 +
<li><strong>i</strong> (insert) Create new files in a directory or copy new files to a directory</li>
 +
<li><strong>d</strong> (delete) Remove files and sub-directories from a directory</li>
 +
<li><strong>a</strong> (administer) Change a directory's ACL. The owner of a directory can always change
 +
the ACL of a directory and any subdirectories in that directory</li>
  
Lmod presents a hierarchal view of the available modules. This means that the view of available modules changes depending on the module loaded by the user.
+
</ul>
 +
<li>File</li>
 +
<ul>
 +
<li><strong>r</strong> (read) Read the contents of any file in the directory</li>
 +
<li><strong>w</strong> (write) Modify the contents of files in a directory and use chmod on them</li>
 +
<li><strong>k</strong> (lock) Allows programs to lock files in a directory</li>
 +
</ul>
 +
<li>Groups of groups supported in AuriStorFS</li>
 +
  </ul>
 +
  <li>Machine-based ACLs</li>
 +
  <li>Native Kerberos integration</li>
 +
  <li>Simple enforcement of quotas</li>
 +
  <li>To make a volume available, mount it somewhere, using 1 or 2 commands. It then becomes
 +
      immediately visible to every client accessing the cell</li>
 +
  <li>Reconfigurations with no user impact; AFS client automatically discovers new configuration</li>
 +
  <ul>
 +
<li>Move volumes to a different partition, on same or different file server</li>
 +
<li>Change what volume is mounted at which mountpoint (new volume at existing path)</li>
 +
  </ul>
 +
  <li>On-line backup volumes</li>
 +
  <li>Client caching</li>
 +
  <ul>
 +
  <li>Software is downloaded to local disk, so is local when accessed</li>
 +
  </ul>
 +
  <li>Collaboration via cells across geographic regions</li>
 +
</ul>
  
 +
AuriStorFS (https://www.auristor.com/) is a commercial implementation of
 +
OpenAFS with important enhancements in performance, security, capacities,
 +
authorization, per-file ACLs, and administration relative to OpenAFS.
  
<pre code>
+
[https://www.auristor.com/filesystem/auristor-comparison Filesystem comparison]
  
$ module available
+
[https://indico.cern.ch/event/531810/contributions/2330922/attachments/1358122/2053923/Auristor_Fact_Sheet.pdf Fact sheet]
  
<nowiki>----------------------------------------------------- /opt/site/modulefiles ------------------------------------------------------</nowiki>
+
<li>Location of database servers and file servers : TBD</li>
  cuda/10.0.130        julia/1.0.4    matlab/2019a          polyflow/20.1.0 (D)    python/3.8 (D)
+
  cuda/10.1.243        matlab/2018a    matlab/2020a    (D)    python/3.6            raspa2/2.0
+
  cuda/10.2.89  (D)    matlab/2018b    polyflow/19.4.0        python/3.7            vim/8.2
+
  
<nowiki>-------------------------------------------------- /opt/ohpc/pub/modulefiles ----------------------------------------------------</nowiki>
+
<li>Public cell - i.e., open to the Internet</li>
   EasyBuild/3.9.2      cmake/3.15.4   gnu8/8.3.0              papi/5.7.0    singularity/3.4.1
+
<ul>
  autotools            gnu/5.4.0      intel/19.0.4.243        pmix/2.2.2   valgrind/3.15.0
+
   <li>Cell name : njhpcfed.rutgers.edu (?)</li>
   charliecloud/0.9.7    gnu7/7.3.0      intel/19.1.0.166 (D)    prun/1.3
+
   <li>Cell alias : njhf (?)</li>
 +
  <ul>
 +
  <li>Instead of, e.g., <em>cd /afs/njhpcfed.rutgers.edu/...</em> : <strong><em>cd /afs/njhf/...</em></strong></li>
 +
   </ul>
 +
   <li>Path to everything in the njhpcfed.rutgers.edu cell : /afs/njhf/.....</li>
 +
</ul>
  
</pre>
+
<li>Find out what startup service AuriStor provides for new cells</li>
  
There are 3 versions of gnu compilers and 2 versions of intel compilers. Load the 'gnu8' compiler
+
<li>ARCS would be willing to provide guidance and technical help in the management of the njhpcfed.rutgers.edu cell</li>
  
<pre code>
+
<li>Access various cells, e.g., /afs/grand.central.org, /afs/athena.mit.edu, /afs/cad.njit.edu, ...</li>
$ module load gnu8
+
<ul>
 +
<li>Demonstrate some user and administrative interactions</li>
 +
</ul>
 +
<li>Use cases</li>
  
$ module available
+
Works anywhere that "fopen" does, including all of: RW, RW+, R, sequential or random access (as opposed
 +
to http/https, which is RO sequential only).
  
<nowiki>------------------------------------------------- /opt/ohpc/pub/moduledeps/gnu8 --------------------------------------------------</nowiki>
+
<ul>
   R/3.5.3        impi/2019.4.243        metis/5.1.0      ocr/1.0.1        pdtoolkit/3.25      scotch/6.0.6
+
  <li>Software</li>
   gsl/2.5        impi/2019.6.166 (D)   mpich/3.3.1      openblas/0.3.5   plasma/2.8.0        superlu/5.2.1
+
   <li>Scripts</li>
  hdf5/1.10.5    likwid/4.3.4          mvapich2/2.3.1    openmpi3/3.1.4    py2-numpy/1.15.3
+
   <li>Data repository</li>
 +
   <li>Documentation </li>
 +
   <li>Web content</li>
 +
</ul>
  
<nowiki>----------------------------------------------------- /opt/site/modulefiles ------------------------------------------------------</nowiki>
+
<li>AuriStor Licensing</li>
  cuda/10.0.130        julia/1.0.4     matlab/2019a          polyflow/20.1.0 (D)    python/3.8 (D)
+
<ul>
  cuda/10.1.243        matlab/2018a    matlab/2020a    (D)    python/3.6            raspa2/2.0
+
<li>Basic : $21,000/cell/yr</li>
  cuda/10.2.89  (D)    matlab/2018b    polyflow/19.4.0        python/3.7            vim/8.2
+
        <ul>
 +
          <li> Security and bug fixes, support for current operating systems, new
 +
      features as they become available, and technical support</li>
 +
  <li> Unlimited AuriStor client software for all supported platforms</li>
 +
                  <li> Unlimited support via email and web 9-5 M-F EST with 4-hour response time</li>
 +
                  <li> 4 database servers</li>
 +
                  <li> 4 file servers</li>
 +
                  <li> 1000 user or machine IDs</li>
 +
        </ul>
 +
 +
<li>Optional</li>
 +
<ul>
 +
<li>Additional database server or file server : $2,500/server/yr</li>
 +
<li>On-site consultant : $1.500 for 8 hours, plus travel and hotel</li>
 +
</ul>
 +
</ul>
  
<nowiki>--------------------------------------------------- /opt/ohpc/pub/modulefiles ----------------------------------------------------</nowiki>
+
<li>Database and file servers hosting and cost</li>
  EasyBuild/3.9.2      cmake/3.15.4    gnu8/8.3.0      (L)    papi/5.7.0    singularity/3.4.1
+
  autotools            gnu/5.4.0      intel/19.0.4.243        pmix/2.2.2    valgrind/3.15.0
+
  charliecloud/0.9.7    gnu7/7.3.0      intel/19.1.0.166 (D)    prun/1.3
+
  
</pre>
+
TBD
 
+
</ol>
Notice that various new modules are now available including various MPI distributions.
+
 
+
===module spider===
+
 
+
module spider provides information on the software installed and how to load the modules necessary to run the software.  For example to find netcdf:
+
 
+
<pre code>
+
$ module spider netcdf
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  netcdf:
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
    Description:
+
      C Libraries for the Unidata network Common Data Form
+
 
+
    Versions:
+
        netcdf/4.5.0
+
        netcdf/4.6.1
+
        netcdf/4.6.3
+
    Other possible modules matches:
+
        netcdf-cxx  netcdf-fortran  pnetcdf
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  To find other possible module matches execute:
+
 
+
      $ module -r spider '.*netcdf.*'
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  For detailed information about a specific "netcdf" module (including how to load the modules) use the module's full name.
+
  For example:
+
 
+
    $ module spider netcdf/4.6.3
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
</pre>
+
 
+
Use the full version to get more information.
+
 
+
<pre code>
+
$ module spider netcdf/4.6.3
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  netcdf: netcdf/4.6.3
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
    Description:
+
      C Libraries for the Unidata network Common Data Form
+
 
+
 
+
    You will need to load all module(s) on any one of the lines below before the "netcdf/4.6.3" module is available to load.
+
 
+
      gnu8/8.3.0  impi/2019.4.243
+
      gnu8/8.3.0  impi/2019.6.166
+
      gnu8/8.3.0  mpich/3.3.1
+
      gnu8/8.3.0  mvapich2/2.3.1
+
      gnu8/8.3.0  openmpi3/3.1.4
+
      intel/19.0.4.243  impi/2019.4.243
+
      intel/19.0.4.243  impi/2019.6.166
+
      intel/19.0.4.243  mpich/3.3.1
+
      intel/19.0.4.243  mvapich2/2.3.1
+
      intel/19.1.0.166  impi/2019.4.243
+
      intel/19.1.0.166  impi/2019.6.166
+
      intel/19.1.0.166  mpich/3.3.1
+
      intel/19.1.0.166  mvapich2/2.3.1
+
 
+
    Help:
+
 
+
      This module loads the NetCDF C API built with the gnu8 compiler
+
      toolchain and the mpich MPI stack.
+
 
+
      Note that this build of NetCDF leverages the HDF I/O library and requires linkage
+
      against hdf5. Consequently, the phdf5 package is loaded automatically with this module.
+
      A typical compilation step for C applications requiring NetCDF is as follows:
+
 
+
      $CC -I$NETCDF_INC app.c -L$NETCDF_LIB -lnetcdf -L$HDF5_LIB -lhdf5
+
 
+
      Version 4.6.3
+
</pre>
+
 
+
Following the instructions above , to load netcdf/4.6.3 compiled with openmpi3 based on the gnu8 compilers, the command would be:
+
 
+
<pre code>
+
$ module load gnu8/8.3.0  openmpi3/3.1.4 netcdf/4.6.3
+
 
+
$ module list
+
 
+
Currently Loaded Modules:
+
  1) gnu8/8.3.0  2) openmpi3/3.1.4  3) phdf5/1.10.5  4) netcdf/4.6.3
+
</pre>
+
Note the phd5/1.10.5 is automatically loaded as a dependency.
+
 
+
 
+
It is useful to use regular expressions to search through the hierarchy to find software packages. For example:
+
 
+
<pre code>
+
 
+
$ module -r spider '.*netcdf.*'
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  netcdf:
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
    Description:
+
      C Libraries for the Unidata network Common Data Form
+
 
+
    Versions:
+
        netcdf/4.5.0
+
        netcdf/4.6.1
+
        netcdf/4.6.3
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  For detailed information about a specific "netcdf" module (including how to load the modules) use the module's full name.
+
  For example:
+
 
+
    $ module spider netcdf/4.6.3
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  netcdf-cxx:
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
    Description:
+
      C++ Libraries for the Unidata network Common Data Form
+
 
+
    Versions:
+
        netcdf-cxx/4.3.0
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  For detailed information about a specific "netcdf-cxx" module (including how to load the modules) use the module's full name.
+
  For example:
+
 
+
 
+
    $ module spider netcdf-cxx/4.3.0
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  netcdf-fortran:
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
    Description:
+
      Fortran Libraries for the Unidata network Common Data Form
+
 
+
    Versions:
+
        netcdf-fortran/4.4.4
+
        netcdf-fortran/4.4.5
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  For detailed information about a specific "netcdf-fortran" module (including how to load the modules) use the module's full name.
+
  For example:
+
 
+
    $ module spider netcdf-fortran/4.4.5
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  pnetcdf:
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
    Description:
+
      A Parallel NetCDF library (PnetCDF)
+
 
+
    Versions:
+
        pnetcdf/1.8.1
+
        pnetcdf/1.9.0
+
        pnetcdf/1.12.0
+
 
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
  For detailed information about a specific "pnetcdf" module (including how to load the modules) use the module's full name.
+
  For example:
+
 
+
 
+
    $ module spider pnetcdf/1.12.0
+
<nowiki>------------------------------------------------------------------------------------------------------------------------------</nowiki>
+
 
+
 
+
 
+
</pre>
+

Latest revision as of 16:34, 5 October 2020

Objective: RU, MSU, NJIT technical staffs judge the value that AuriStorFS would likely bring to the NJHPCFED, and later to the Eastern Regional Research Platform (ERRP).

If that value is considered sufficient, proceed with budgeting the purchase of AuriStorFS.

  1. Go over in detail the items listed below
    • Designed for wide area network
    • Single global namespace for all clients. All clients see exactly the same path and ACLs.
    • Clients for all platforms - Linux, MacOS X, Windows, others
    • Extremely efficient administration and applications distribution
      • Administration by any authorized Kerberos principal from any AFS client
    • Long history of working well and reliably at many institutions and international corporations
    • No administrator intervention in making mount points available to all clients, other than creating the mount point - a single command done from any AFS client
    • Volume : basic storage unit. Volumes work very well as specific-purpose containers, e.g., software, datasets, web content, documentation)
    • Read-only replication of volumes
    • Scalability - number of clients, volumes, users are readily accommodated
    • Fine-grained ACLs CMU doc
      • Directory
        • l (lookup) List the contents of a directory; does not allow the reading of files
        • i (insert) Create new files in a directory or copy new files to a directory
        • d (delete) Remove files and sub-directories from a directory
        • a (administer) Change a directory's ACL. The owner of a directory can always change the ACL of a directory and any subdirectories in that directory
      • File
        • r (read) Read the contents of any file in the directory
        • w (write) Modify the contents of files in a directory and use chmod on them
        • k (lock) Allows programs to lock files in a directory
      • Groups of groups supported in AuriStorFS
    • Machine-based ACLs
    • Native Kerberos integration
    • Simple enforcement of quotas
    • To make a volume available, mount it somewhere, using 1 or 2 commands. It then becomes immediately visible to every client accessing the cell
    • Reconfigurations with no user impact; AFS client automatically discovers new configuration
      • Move volumes to a different partition, on same or different file server
      • Change what volume is mounted at which mountpoint (new volume at existing path)
    • On-line backup volumes
    • Client caching
      • Software is downloaded to local disk, so is local when accessed
    • Collaboration via cells across geographic regions

    AuriStorFS (https://www.auristor.com/) is a commercial implementation of OpenAFS with important enhancements in performance, security, capacities, authorization, per-file ACLs, and administration relative to OpenAFS.

    Filesystem comparison

    Fact sheet

  2. Location of database servers and file servers : TBD
  3. Public cell - i.e., open to the Internet
    • Cell name : njhpcfed.rutgers.edu (?)
    • Cell alias : njhf (?)
      • Instead of, e.g., cd /afs/njhpcfed.rutgers.edu/... : cd /afs/njhf/...
    • Path to everything in the njhpcfed.rutgers.edu cell : /afs/njhf/.....
  4. Find out what startup service AuriStor provides for new cells
  5. ARCS would be willing to provide guidance and technical help in the management of the njhpcfed.rutgers.edu cell
  6. Access various cells, e.g., /afs/grand.central.org, /afs/athena.mit.edu, /afs/cad.njit.edu, ...
    • Demonstrate some user and administrative interactions
  7. Use cases
  8. Works anywhere that "fopen" does, including all of: RW, RW+, R, sequential or random access (as opposed to http/https, which is RO sequential only).

    • Software
    • Scripts
    • Data repository
    • Documentation
    • Web content
  9. AuriStor Licensing
    • Basic : $21,000/cell/yr
      • Security and bug fixes, support for current operating systems, new features as they become available, and technical support
      • Unlimited AuriStor client software for all supported platforms
      • Unlimited support via email and web 9-5 M-F EST with 4-hour response time
      • 4 database servers
      • 4 file servers
      • 1000 user or machine IDs
    • Optional
      • Additional database server or file server : $2,500/server/yr
      • On-site consultant : $1.500 for 8 hours, plus travel and hotel
  10. Database and file servers hosting and cost
  11. TBD