Step 1: Log In
Once a new user has obtained an HPCVL username they will be able to
login on the computer of their choice.
Use the following hostname to login to each machine.
SunFire 6800
|
sfnode0.hpcvl.carleton.ca
|
Beowulf Cluster
|
thog.hpcvl.carleton.ca
|
M4000 Server
|
carlm4k.hpcvl.carleton.ca
|
In order for any HPCVL users to gain access to
resources at Carleton, they must use Secure Shell (SSH).
To create a SSH connection to one of the machines,
a SSH client software is required.
On Linux and Mac OS, a SSH client is usually installed by default.
To connect to one of the clusters,
- on Linux and Mac, run the following command and enter your password when prompted
ssh username@hostname
on Windows platform, use the following instruction
Putty Secure Shell Client
username is your HPCVL account username.
hostname could be obtained from the table above.
For more information about Secure Shell please
consult the HPCVL SSH
FAQ [hpcvl.org].
Should you experience difficulties logging in,
contact the HPCVL-Carleton System Administrator.
Step 2: File Transfer
In order to transfer files to/from one of the clusters from/to a local machine,
HPCVL users should use either Secure Copy (SCP) or Secure File Transfer Protocol (SFTP).
To transfer files to one of the clusters, on Linux and Mac,
-
scp ./mydata.dat username@hostname:myfile.dat
will copy "mydata.dat" file from the current directory on your local machine
to the home directory on the hostname machine.
-
scp -r myfolder username@hostname:mydata/
will copy directory "myfolder" recursively (i.e., the directory and all its contents) from
your local machine into directory "mydata" in your home directory on the hostname machine.
To transfer files from one of the clusters, on Linux and Mac,
-
scp username@hostname:mydata.dat ./myfile.dat
will copy file "mydata.dat" from the home directory on your local machine to
your current working directory on the hostname machine.
-
scp -r username@hostname:mydata ./myfolder
will copy directory "mydata" recursively (i.e., the directory and all its contents) from
the hostname machine into directory "myfolder" in your current working directory on your local machine.
provide your password when prompted.
username is your HPCVL account username.
hostname could be obtained from the table above.
To transfer files to/from one of the clusters, on Windows platform,
use instructions on using WinSCP SFTP/SCP Client.
For more information about transferring files to/from the clusters please
consult the HPCVL SSH
FAQ [hpcvl.org].
Should you experience difficulties transferring your files to/from one of the clusters,
contact the HPCVL-Carleton System Administrator.
Step 3: Submitting Jobs
In order to properly manage system resources and ensure that all HPCVL
users have fair and equitable use of HPCVL computing facilities jobs
should be submitted through a job scheduler.
To manage all of HPCVL Carleton's computing resources Sun
Microsystem's Grid Engine is used.
These are instructions to help you get started with a sample
GridEngine job on the HPCVL-Carleton cluster. Once this is working,
you should be able to modify it for your own needs.
- Create an MPI Program
If you do not already have your own MPI program, you may use
our sample MPI program in your
own home directory. For example:
Log into thog.hpcvl.carleton.ca using ssh
Download the example program by running:
wget http://people.scs.carleton.ca/~hpcvl/mpi_example.tar.gz
Uncompress the example by running:
tar zxvf mpi_example.tar.gz
Comile the program by running:
cd mpi_example
make
- Create an SGE Script
In your mpi_example directory, there should already be an
example.sge script.
- Run SGE
Submit the mpi_example job to GridEngine by running:
qsub -pe mpi 8 example.sge
- Check your job's status
Check on unfinished jobs by running (your job will not be
displayed after it is finished):
qstat
- View your job's output
When your job has finished, the output should be in
mpi_example/hello_output.txt
For more SGE information visit HPCVL-Carleton Gridengine Guide.
Users who are unfamiliar with the
proper use of Grid Engine should consult the HPCVL
Grid Engine FAQ [hpcvl.org].
Step 4: Set Up Your Environment
Once a new user is logged on to one of the systems the first thing
they want to do is set up their user environment. Each user has their
own preferences on which shell to use and how they would like to
customize their system settings. Often users like to use different
shells, command line editors, and development tools. By setting up
their user environment a new user can work in an environment that is
both familiar and easy to use.
For more information on setting up your user environment
on any of HPCVL's machines, please consult the HPCVL
Setup FAQ [hpcvl.org].
Step 5: Develop Your Program
It is the expectation that all new users to HPCVL should have some familiarity
with developing program code. HPCVL users who are new to parallel program
development can obtain more information at the HPCVL Parallel Programming FAQ [hpcvl.org].
Step 6: Save Your Data
As programs are running they sometimes use or generate large amounts of
data. In conventional computing environments system resources can become
quickly overwhelmed as some jobs start synthesizing data and storing
results. Users of HPCVL Carleton's computing facilities are required to adhere
to policies and procedures outlining the proper use and management of system
storage and archive facilities. These policies have been put in place to
ensure that all HPCVL users have access to adequate and reliable storage for
all their research needs. All HPCVL Carleton users are required to adhere to
these policies as they help protect the integrity of experiment data and
research results.
|