[GE users] sge questions

Forum Parmar forum.parmar at gmail.com
Thu May 31 00:17:58 BST 2007


    [ The following text is in the "ISO-8859-1" character set. ]
    [ Your display is set for the "ISO-8859-10" character set.  ]
    [ Some special characters may be displayed incorrectly. ]

Hi,
I have just started to use SGE and had a couple of questions about the
engine which I did not really understand after going through the manual.

Assuming that SGE is installed for a small cluster:

1a. If I have hosts with different resources (like memory, number of CPUs,
number of cores etc) how does the engine know of these? Do I have to do
something to make it aware.

1b. Is a job assigned to a
      i) host or
      ii) cpu or
      iii) core on a host?

2. In what cases is a job divided and run over multiple hosts/cpus/cores?
2a. How does it work if I know the application is multithreaded?
2b. What if a job is really big and not multithreaded, how will it be run on
the cluster?

3a. If a job requires to read from an input file, and assuming that job is
run on multiple hosts, how does every host get access to that input file? (I
understand that the the SGE_STDIN_PATH environment variable must be set with
the qsub command but not sure about multithreaded cases).

3b. Also,  Where does the output go in that case? Is the output combined
into one output file automatically?

Thanks in advance,
Forum



More information about the gridengine-users mailing list