These pages explain how to utilize high
performance computing (HPC) at NYU.
To get started, you'll need to request
an account on the system. As part of the process,
you must choose a shell. This
is your text-based interface with the system. I
recommend choosing "bash," which is the default, I
believe, unless you are an expert with a different
shell. A professor must authorize your account. You
will need to renew your account once a year, and it's
best to have your advisor be your sponsor. Your NYU
login ID (and its associated password) is your login on
the system. There are two systems, "Union Square" and
"Bowery." The latter should be used for parallel jobs,
such as running the model. Analysis (matlab, etc.)
should be done on Union Square, or you can bring key
data to a Courant machine for local analysis. To log
on, connect to the gateway via SSH.
(Note that the -Y allows you to tunnel x-windows, the -l
lets you specify your username.)
ssh -Y -l nyuID hpc.es.its.nyu.edu
where nyuID should be your login ID. Then, connect to union square or bowery, respectively:
ssh -Y usq.es.its.nyu.edu
or
ssh -Y bowery.es.its.nyu.edu
The HPC is a
linux system. If you're not familiar with linux, I
recommend browsing the web for a tutorial. CIMS
has some pages on linux, and this online
tutorial looks pretty reasonable. (The tutorial is for
unix, but linux is a unix-like system, and all the basic
commands are the same.)
The following PDFs explain how to get started with the model.
If you have any comments or questions, please write me, so I can
improve them for future users!
Setting up the model in
your HPC account.
How to first run the model, using an interactive session.
How to use the queue for production runs.
Basics on analyzing the model output.
A handy matlab script to read date from netcdf files.
How to compile the model
(necessary for changing parameters such as the rotation rate or radius
of the Earth.)
Trouble shooting: What to do if the model doesn't run?
Linux help and shortcuts (a work in progress).
A brief, incomplete introduction to using postproccessing scripts to speed up your analysis.