How to Access the Instructional HPC Clusters

The CoC provides two instructional HPC clusters, deepthought and jinx. SSH access to the instructional HPC clusters is provided through it's head node, named <cluster>-login.cc.gatech.edu..  You must log into the head node with your Georgia Tech user name and password. The head node has limited compute and memory capacity therefore compiling and running jobs should be done on compute nodes instead.  Jobs can be submitted using the qsub command on jinx or sbatch on deepthought from the cluster head node.

Once logged into <cluster>-login, you can compile programs on compute nodes using the interactive feature of the qsub/salloc command and subsequently submit jobs to the compute nodes through the various Torque and Slurm queues.  Submitting a job with the qsub/sbatch command is the only way to run programs on the compute nodes; you cannot log into nodes that are not currently allocated to a job you own.

Submitting Jobs

For information on running compute jobs on the Jinx cluster, see How to Run Jobs on the Instructional HPC Clusters.     Note:  Since <cluster>-login is shared by many users, please do not attempt to run resource-intensive programs on the head node, even for compiling/testing/debugging.  You should submit your job to the appropriate Torque or Slurm queue.