Just brainstorming, here are some ideas as to why it might be faster:
- The machines in your Condor pool might simply have faster
cpu/memory/disk
than your desktop machine. Trying logging directly into one and running
your job.
Tried that... I am running condor on a lab of computers at Charles Sturt
Uni, Australia... all the computers are the same... If I run it on one of
those computers NOT through condor... my job takes about 4 minutes to
run.... If I run it through condor... about 1 minute!!!! Now that is really
good... but it just doesn't make sense!!!
- If you job produces a lot of output, it will run considerably faster
sending its output to a file than to the terminal. (Scrolling a big
graphic display can be expensive.) Try timing both of those cases.
I figured this might be the case... so I ran the same job but took out all
the printing to screen...... same result... 4 minutes out of condor one
minute with condor.
- On your desktop machine, you might be writing your output to a
distributed
file system, which can be considerably slower than using a local disk.
For example, if your /home directory is stored on NFS and /tmp is a local
disk, then "myprog > /home/bob/output" can be considerably slower than
"myprog > /tmp/output".
Nope... that is defianatly not the case... all running of the local disk
(Windows XP).
Those are just some ideas...
Let us know what you discover.
Thanks Doug... I'll just have to keep guessing.... but at the end of the
day... I'm not going to stress about it.... It's running my job 4 times as
fast :)
Cheers.
Condor Support Information:
http://www.cs.wisc.edu/condor/condor-support/
To Unsubscribe, send mail to majordomo@xxxxxxxxxxx with
unsubscribe condor-users <your_email_address>
![[Computer Systems Lab]](http://www.cs.wisc.edu/pics/csl_logo.gif){kind=logo}