FAQ:
Could you tell me of a good computer to use?
ANSWER:
Your supervisor or project lead are the best source of information about the resources available. Most projects have their own record of which computers should be used for which tasks. For example:
General Purpose Interactive Computing
lnx201.classe.cornell.edu
is CLASSE's general purpose remote login node. It can be used for light interactive jobs, but should not be employed for serious compute jobs, since that could lock other users out.
Building code
In general, you cannot assume that your desktop computer will be able to build usable executables for processing on other nodes. The operating systems may be incompatible. For example, CLEOns
must compile and link on lnx134, which runs an older compiler compatible with the CLEO libraries. You can compile and link on the target node, or ask your supervisor or colleagues.
Running batch jobs
It is strongly preferred to run compute-intensive jobs in the batch queues, where the nodes have the memory, fast CPUs, and direct access to disk farms, and where queuing software apportions usage fairly and systematically. Although nearly all the batch nodes were bought by CesrTA, CLEO, and CMS for their use, anyone in CLASSE can use them in moderation. Researchers needing large amounts of CPU time (say, more than 5 CPU days) should contact David Kreinick (
dlk8@cornell.edu) to make sure they won't be interfering with CLEO or CesrTA production jobs. You can submit jobs to the Linux farm using
GridEngine.
Running interactively
It is possible to reserve a batch queue slot for interactive computing; see the qrsh command in
GridEngine. As mentioned above, light jobs can be run on lnx201. DO NOT log into any other node interactively unless you are sure its assigned user is not trying to get work done on it. It is easy to tie up a desktop's resources, making it unusable for the person sitting at the console.
NFS file systems
A list of available NFS filesystems is available from
NfsDirs.
Back to: ComputingFAQ