Machine Names | photon00 to photon30 |
Number of nodes | 31 |
Current status | photon[14,20,21,26,30] DOWN |
Hardware | Dual Xeon 2.4GHz, 2GB Memory per node. (4GB for node 24,25,27,28) |
OS | CSAIL Debian Linux |
File System | CSAIL AFS (mount to Galactus NFS available upon request) |
Everybody in our group have access to the cluster, and typically the usage has been low during non-peak times (i.e. >4 weeks before SIGGRAPH). So feel free to use the nodes to do your heavy computation task. However, please use top/jmon to check machine load before launching your process, to make sure you are not bogging down an already busy node. It is frustrating to be bottlenecked by one of the 20 parallel jobs you are running. On a related note, if you are distributing your processes over many nodes, be nice and keep a couple of the nodes free.
Each node can be accessed through SSH directly. On Windows, if you are running Kerberos, you can also configure your SSH client to forward the kerberos directly, so that you don't need to type your password (this applies to other CSAIL machines as well).
** First make sure you add the line 'matlab' into your ~/.software file. This will add the default matlab path to your environment.
To run matlab without a prompt (e.g. if you want to run it as a background process, or you just hate it like Daniel), create a shell script runmatlab.sh like this:
#!/bin/csh
/afs/csail.mit.edu/system/i386_linux24/matlab/latest/bin/matlab -nodesktop
-nosplash << TAG
myfun('$1');
exit
TAG
and replace the code in red with the .m code you want to execute. In the example I showed, $1 take the first argument of the shell script and pass it to matlab. Now you can run batch job on remote nodes like this:
rsh -n photon00 runmatlab.sh crap0 &
rsh -n photon01 runmatlab.sh crap1 &
rsh -n photon02 runmatlab.sh crap2 &
.
.
.
(PS and then you will find out we only have a limited number of licenses for each Matlab package)
A: For some reason, /afs/csail.mit.edu/system/i386_linux24/matlab/latest/ is pointing to version 6.5. Newer versions including R14sp1 and R14sp2 are available at /afs/csail.mit.edu/system/i386_linux24/matlab/.
For usage, how-tos, ask Addy (or Paul or Soonmin). For hardware problems, software installation requests, contact Brian Jones from TIG.