Good Practice Guidelines

1) Use your data area for job input and output and not your home area.

The home area can be used by jobs if their I/O usage is light and the amount of data being accessed  is relatively small.

Please note ATHENA users : ATHENA data and ATHENA itself can take up a lot of space.  It is preferable that you install and run it from your data area.

Data areas are of the form "/data/dataXXX/USERNAME".

All users are entitled to have access to their own data area if they require it. Please contact system administrator for more details.

If you have run data on your home area, please move it to your data area.

There are a number of reasons why using your data area over your home area is recommended.

Faster run time - The data disk servers by design are geared for speed and handling multiple accesses as would be incurred by dozens of simultaneous jobs.
The home area disk setup by design is not this rigorous. Also, the home areas are located on a separate subnet from the cluster and as a result, cluster jobs accessing home areas instead of data areas are subject to higher network latencies and can result in bottlenecks.

Access to more disk space - As a cluster user, you have access to much more space on your data drive than on your home area.  This is because the data drives themselves are quite large.  The home areas are kept smaller by design to facilitate the nightly backup.  The data areas are not backed up due to their large size, but have higher reliability due to the RAID technology they employ.

More socially responsible - Since home areas are typically smaller and shared among a number of users, if you use your home area for data storage for cluster jobs, you can easily fill up the disk and cause serious problems for other users on your partition.  This scenario should be obviously be avoided.  Also, if you have multiple cluster jobs hammering the home area disks, this could could cause network, home server or other problems on the home server side which could negatively affect others and get people upset.

2) Good Scripting Practice - Use variables instead of absolute values for data (and other) directory specification

 Due to the ever changing requirements of data disk users, occasionally it is necessary to move a user's data area to another partition.

For this reason, it's in the user's best interest to use variables for any file specifications used in any user scripts.  This way, whenever a parition change is required or the script for any reason is ported to run in a different location, the change can be done easily. ie on one line

For example, in this csh script :

set backdir='/data/data004/backup_store1'
set temp='/tmp'
echo " File Backup  . . . "
/local/bin/tar cvf $temp/user1.tar /home/user1
echo " Moving files  . . . "
mv $temp/user1.tar $backdir

3) Inform others when if you intend to ramp up your usage

If you intend to use more than fifty high priority queues, please inform all others in your submission group.  Otherwise, you might be hogging all the queues at a time when cluster usage is critical for others.
The queuing system itself can be modified to partially address situations like these but since user requirements change so drastically and so quickly, its best just to let people know.

For more information on running jobs on the Physics compute clusters, have a look at :
Research Compute Cluster