Hi,

The master node is 
        molly0.ling.uni-potsdam.de
(also known as emma.ling.uni-potsdam.de)

and the slave nodes are
        molly1.ling.uni-potsdam.de
        molly2.ling.uni-potsdam.de
        molly3.ling.uni-potsdam.de
        molly4.ling.uni-potsdam.de

your login name is the same as on helios, your password will be sent
with a separate mail. You will have to change your password on all nodes
separately. You might want to follow the hints in [1]. ClusterSSH ([2]) 
is a useful tool to access all cluster nodes (or a subset) in parallel.

Your home directory is mounted from helios as well as /projekte. 

We use Sun Grid Engine as our batch job scheduling system. From any 
node, you can submit jobs to the cluster (qsub) , you can open shells 
that will automatically use the least loaded node (qlogin) and there is 
a parallel make (qmake). A very thorough introduction is available in 
[3], also note the manual pages: try "man sge_intro" as well as the 
man-pages for qsub, qlogin and qmake.

I understand that for now, we will just continue to login to different
nodes and start our batch jobs with whatever scripts we've used before,
but we should probably try to standardize on SGE for long-running jobs
in order for scheduling policies to be effective. 

Each cluster node has 4 to 8 processor cores, so you can increase
performance if you (a) run processes in parallel or (b) parallelize your
program (for example with multithreading). Each machine has a total of 
16GB of memory, but due to the architecture, only about 8 GB can be 
effectively used by one process. Molly4, the 4 processor machine has 
faster CPUs and more memory per CPU, so you may want to run 
time-critical tasks there.

We do not enforce limits to your resource usage so please be a good 
citizen when sharing machines with other users, both in terms of CPU 
cycles and memory.

You may notice your performance to be IO bound, when intensively 
accessing data from /home and /projekte. Data intensive tasks (like copy 
operations) may be faster on helios. You might also want to copy often 
used read-only data to /tmp once and then access it locally. Note that 
/tmp is not shared between nodes.

There is hardly any X (aside from emacs) on the nodes and most
software will be missing. Just tell us what you need and we are likely 
to install it.

Updates to this information will be posted to [4].

Cheers,
Timo

[1]: http://www.schlittermann.de/doc/ssh.html
[2]: http://clusterssh.wiki.sourceforge.net/Main+Page
[3]: /opt/pkg/sge/doc/SGE6UserDoc.pdf
[4]: http://www.ling.uni-potsdam.de/~timo/cluster