Linux User Instructions
Shale User Accounts
Contact Aaron Bergstrom to receive a Shale User Account. His contact information is listed at the bottom of this page. Use SSH, SCP, SFTP to connect to the shale cluster. Typical SSH login attempt would look like the following:
Example: ssh email@example.com
Production Cluster vs Test Cluster
Users should use the Shale Test Cluster when writing, compiling, and testing their code. To access the test cluster, you must login the the test cluster admin node. You do so after you've already logged into Shale by typing the following on the command line:
Example: ssh shalecustom1
Test jobs are submitted from this node in the same manner as they are on the production cluster.
To return to the production cluster, type either "exit" or use ssh from the command line to move to any Shale production cluster admin node.
Example: ssh shale1
Shale Production Cluster Nodes
There are three different types of nodes within the Shale Production cluster: four Administration nodes, 24 Computation nodes, and four CUDA nodes. They are divided up by the following sections:
shale1, shale2, shale3, shale4
node0, node1, node2, node3, node4, node5,
node6, node7, node8, node9, node10, node11,
node12, node13, node14, node15, node16, node17
node18, node19, node20, node21, node22, node23
node24, node25, node26, node27*
*Currently, nodes 25 and 27 aren't compiling CUDA correctly.
To access any of the admin nodes, you must use the ssh command, for example:
To access any of the computation or CUDA nodes simply just type the name of the node, for example: node24
Shale Test Cluster Nodes
There are two different types of nodes within the Shale Test cluster: an Administrative node and eight Computation nodes. They are divided up by the following sections:
testnode1, testnode2, testnode3, testnode4, testnode5, testnode6, testnode7, testnode8
To access shalecustom1, you must use the ssh command, for example:
To access any of the computation nodes simply just type the name of the node, for example: testnode3
Shale User Paths
Instructions for adding compiler path information to your enviornment can be found on the Shale User Paths page .
PBS Submission Scripts
Instructions for PBS Submission Scripts and their usage, can be found on the Shale Submission Script page .
Once you have setup your path information as described above, do the following to compile C, C++, Fortran code:
For C: mpicc sourcecode.c -o executable.x
Where sourcecode.c is the name of the C source code file, and executable.x is the name of the executable file to be created.
For C++: mpixx sourcecode.cpp -o executable.x
Where sourcecod.cpp is the name fo the C++ source code file, and executable.x is the name of the executable file to be created.
See the CUDA Programs page for directions on how to compile and run CUDA programs on the Shale Production Cluster.
Executing Your MPI Program
You may use mpirun or mpiexec to execute your compiled mpi program.
Example: mpiexec -n 16 ./executable.x
Where -n is the flag that designates the number of processors and 16 is the number that you wish to request.
Shale is a shared resource. All MPI programs must be executed through the Shale job scheduler. Abuse of this policy may get users banned from the using the system. Instructions for using the job scheduler to execute code may be found on the Shale Submission Script page .