[Condor-users] Questions about NFS and transfer system mechanism

[Nicolas GUIOT <nicolas.guiot@xxxxxxx>]

Hi,

On all of the machines (submit, execute, etc..), we have NFS installed, so that anywhere you can refer to /nfs/my_file. That's the way we call the binaries, and the data files, for read, and write, for example : 

/nfs/bin/my_prog /nfs/my_data_file 
(all this into a correct condor submit file, of course)


So basically, we don't have any problem to read/write for the computing jobs.

Now I have some questions (first, consider I'm quite newbie with computing internals): 
- how does it work for the binary program : is that copied onto the local /scratch, or directly loaded in RAM: Or : what is the best solution in terms of performance, and how can I apply it ?

- About data files : it seems the file transfer mechanism is only used with non-NFS system : So, if I understood well, it means that during the computations, the program accesses the remote file to read/write each time it's necessary : do I _have_ to do it this way, or would it be possible to copy all the data to a local /scratch repertory, then copy everything back at the end ?

Thanks in advance
Nicolas


-----------------------------------------------
CNRS - UPR 9080 : Laboratoire de Biochimie Theorique
Institut de Biologie Physico-Chimique
13 rue Pierre et Marie Curie
75005 PARIS - FRANCE

Tel : +33 158 41 51 70
Fax : +33 158 41 50 26
------------------------------------------------