Prism Aurora and Anasazi

Contents

Accessing the Machine

SSH

You should log in to prism, aurora and anasazi with X11-forwarding via:
[code sh]
ssh -Y [uid]@prism.multiscale.utah.edu
ssh -Y [uid]@aurora.multiscale.utah.edu

ssh -Y [uid]@anasazi.multiscale.utah.edu

[/code]
to be able to run any windowed application like totalview and matlab.

Note that from off-campus you will need to use a VPN connection. For your VPN username, use [uid]@chpc.utah.edu to get the CHPC VPN pool.

FastX

This allows you to access the machines using xwindows through a web browser:

Using the Desktop Client

  1. Download the desktop client.  To do this, point a browser to the appropriate machine (see links above) and rather than logging in, click the link “Looking for the desktop client?” to download it.
  2. Configure the client.  Launch the FastX client and click the “+” sign to add a new connection. Enter [host].multiscale.utah.edu as the host (substituting, prism, aurora or anasazi for host as appropriate) and your UNID as “User”
  3. Connect to the prism connection you just made.
    • Note that you will need a VPN connection in order to connect through the desktop client if you are connecting from off campus.
  4. Click “+”  and then choose “xfce” under the “Suggestions” drop-down menu.  This will launch a xterm desktop environment.  Alternatively, you can select a previous session or launch a specific command rather than clicking the “+” icon.

More Information

Additional information on FastX can be found on the CPHC web page.

Fair Use Policy

These machines are primarily intended for development. If you need to run small cases, feel free, but do not overload the machine.

Also note that aurora runs our Jenkins build/test system, so it will have periodic load spikes associated with code commits.  So you should never load aurora more than 50% for extended periods of time.

Machine Description

prism is a linux desktop machine with the following specs:

  • 2 x Intel Xeon 6 Core E5-2620 2.0 GHz Processors
  • 15 MB L3 Cache
  • 16GB DDR3-1333 PC3-10600 CL9 ECC Registered DDR3 RAM
  • nVidia Tesla K20 with compute capability 3.5.

aurora is a linux desktop machine with the following specs:

  • 2 x Intel Xeon 12 Core E5-2670 v3 2.30GHz processors
  • 30 MB L3 Cache
  • 64 GB RAM
  • nVidia Tesla K80 GPU with 24 GB memory

anasazi is a linux desktop machine with the following specs:

  • 2 x 6-core Intel Xeon E5-2620 2.0 GHz processors
  • 15 MB L3 Cache
  • 32 GB RAM (DDR3-1333)
  • nVidia GeForce GTX-680 GPU card, compute capability 3.0.
  • 5 TB disk space

Filesystems

prism, aurora and anasazi mount several filesystems:

  • CHPC home directory. When you log into prism, you will land in your CHPC home directory. Note that this is SLOW read/write access. You should not run cases or compile code from this directory.

Prism’s specific filesystems

  • 670 GB SSD mounted at 
      /scratch/local/prism_fast

    Please only use this for compiling and not storing datasets. This is FAST disk space, but isn’t large. Samba mount this via:

    smb://ad;[UNID]@prism.multiscale.utah.edu/prism_fast

    and when prompted for your user name, use:

    ad\[UNID]
  • 1TB 7200rpm SATA disk mounted at /scratch/local/prism_large Samba mount this via: 
      smb://ad;[UNID]@prism.multiscale.utah.edu/prism_large

    and when prompted for your user name, use:

      ad\[UNID]
  • The aurora filesystem is cross-mounted on prism at /scratch/remote/aurora

Mounting Prism’s Filesystem

Samba mount via: and your campus password.

On BigMac, prism’s disk is mounted at

  /scratch/remote/prism_fast
  /scratch/remote/prism_large

On aurora, prism’s disks are mounted at

  /scratch/remote/prism_fast
  /scratch/remote/prism_large

Aurora’s Filesystems

  • 700 GB of local SSD scratch disk space located at
  •   /scratch/local/aurora_fast

    This is not backed up, but is persistent (will not be purged) unless you abuse it.

  • 5.5 TB of local scratch disk space located at 
      /scratch/local/aurora_large

    This is not backed up, but is persistent (will not be purged) unless you abuse it.

  • 1 TB NFS-mounted space on prism 
      /scratch/remote/prism_large
  • 700 GB SSD on prism: 
      /scratch/remove/prism_fast

    Note that access to this disk is limited by the network, and its performance will not be fast unless you are on prism itself!

Mounting Aurora’s Filesystem

Samba mount via:

  smb://ad;<UNID>@aurora.multiscale.utah.edu/Aurora

and when prompted for your user name, use:

  ad\[UNID]

and your campus password.

On BigMac, aurora’s disk is mounted at

  /scratch/aurora

If you are connected via Uconnect Wifi, make sure that you set the following username for Uconnect:

  <UNID>@chpc.utah.edu

your password should be the same as your campus password.

Anasazi

Anasazi’s filesystems are essentially the same as prism and aurora, except for their size.

Software installed

It has most software pre-installed. If you need something that is not installed, contact James.

Some software requires you to load modules:

  • totalview: module load totalview
  • matlab: module load matlab

A recent version of visit is available at:

  • visit: /usr/local/visit/bin/visit

Some useful “productivity software”

  • LyX (just run lyx& from a command prompt).
  • LibreOffice: libreoffice&
  • xpdf (pdf reader)
  • inkscape for drawing graphics

Cantera

Many of our codes have a dependency on Cantera.  We maintain our own version that has a few modifications to the main cantera development branch.  This has been installed on prism, aurora and anasazi in [code]/usr/local[/code].  Be sure to set the following environment variables:

[code]

setenv PYTHONPATH /usr/local/lib/python2.7/site-packages

[/code]

This is required to convert .cti files to .xml files as cantera starts up.

If you are having trouble with Cantera on prism or aurora, look at

[code]

/usr/local/cantera/bin/setup_cantera

[/code]

which has a number of environment variables that you may want to set.  In fact, it is best to add:

[code]

source /usr/local/cantera/bin/setup_cantera

[code]

to your shell login script.

Building Uintah

Many of you are interested in using aurora to build Uintah. Here’s one approach to do so:

  • Create a working directory for yourself under /scratch/local/prism_fast or /scratch/local/aurora_fast. It is best that you call this directory by the first letter of your first name followed by your last name. Lets call this jdoe.
  • cd /scratch/local/prism_large/jdoe
  • checkout a fresh version of uintah: git clone https://github.com/Uintah/Uintah.git uintah. This will checkout the latest copy of Uintah and store it under ./uintah
  • change directories to uintah: cd uintah
  • we will start by building an optimized version of uintah: mkdir opt
  • change directories to the optimized folder: cd opt
  • create a configure script: vi uintah-configure-opt.line
  • paste the following:
    [shell]../src/configure
    –with-hypre=/usr/local/hypre_2.11.2
    –with-mpi-include=/usr/include/mpich \
    –with-mpi-lib=/usr/lib/mpich/lib \
    –enable-wasatch_3p \
    –enable-wasatch \
    –without-fortran \
    –with-boost=/usr \
    –enable-64bit \
    –enable-optimize=’-O3′[/shell]

    • make your configure line executable: chmod +x uintah-configure-opt.line
    • execute your configure line: ./uintah-configure-opt.line
    • wait… if unsuccessful, you will see a lot of errors 🙂
    • now build! make -j 10
    • To build in debug, create a directory called dbg and a new configure line uintah-configure-dbg.line.
    • You will need to replace --enable-optimize='-O3' with --enable-debug and you’re good to go.

0 Comments

Leave a Reply

Avatar placeholder