To access ark cluster run the following command:

ssh ark28

Windows user: please read How can I SSH from Microsoft Windows?
When prompted for password use your TSRI password. Contact sargis@scripps.edu to request an account if needed.
If you are planning to use this often, consider setting up SSH without password. Here is how: run these commands from your local machine.

ssh-keygen -t rsa 
scp ~/.ssh/id_rsa.pub ark28:~/.ssh/authorized_keys

Running Jobs

I have installed TORQUE on ark which can be used to run jobs on the cluster. There is plenty of documentation available on Portable Batch System so here I'll cover details relevant for ark cluster. First of all, the /home folder on each node resides on a local hard drive which makes your home folder different on different nodes. Second, pbs clients are running on all nodes but ark28, which runs pbs server.

Administrative Notes

I've found an interesting article To Protect and Serve: Providing Data to Your Cluster and a link to Open Source Cluster Application Resources (OSCAR) in there. Since I already have PBS and Ganglia installed on ark, we don't need OSCAR yet. There is Cluster Command Control (C3) and System Installation Suite (SIS) distributed by OSCAR that seem very handy tools to have, but lets concentrate on running AutoDock on ark for now.

Directory Structure

The /usr/local and /home folders from ark28 are mounted on all nodes under /ark using autofs in read-only and read-write mode, respectively. Note that autofs does on-demand mounting so 'ls /ark' might not show home or local folders but, for instance, 'ls /ark/home' will. I have mounted /usr/local is that I can install binaries on ark28 and make them available on all nodes. When installing AutoDock for the first time, I used a script similar to this one:

for a in `cat hosts`; do echo $a; scp autodock4 root@$a:/usr/local/bin/autodock4; done

Here, host is a file under /root folder that has the names of all hosts in the cluster. While it is true that calling autodock from /usr/local/bin is faster than calling it from /ark/local/bin, installing, maintaining and updating AuoDock or any other software is easier on a single machine (ark28) than installing them on all the nodes.
Similarly, /home from ark28 (master node) is available on all compute nodes. At the end of each job users can copy files created under /home to some place under /ark/home. The /home folder on each machine resides on local hard drive and can be used as a scratch. When user is logged into a compute node with 'qsub -I', for instance, their starting location is under /home. In order to copy files from compute nodes to master node one can use scp, however this requires setting up scp to work without prompting for password that might be time consuming for large cluster. That's why /home is mounted under /ark so that we can use a simple cp command instead.