Besides from having your personal account on The Digital Research Alliance of Canada (the Alliance) - previously and in most of these wiki pages still called "Compute Canada" - you are also member of at least one, probably several "project accounts". These project accounts can be used for two things:


If you do not know of which project accounts you are a member you can execute the command:

    id


Naming convention

All project names start with one of the following:

    def  :  default account
    rrg  :  resources for research groups
    ctb  : contributed

Followed by a '-' and the name or name acronym of the professor.

1) Data storage

To see under which project accounts you can store data check the list of links under ~/projects:

    ls -l ~/projects

To check the space available for each account you are a member of you can use the official command:

    diskusage_report

or my version of the same command which adds also the usage in percent:

    ~winger/ovbin/k.quota

2) CPU allocation

To submit any program on any machine of the Alliance you need to specify which project account should get billed for the use of the cpu time. The project accounts under which one can store data are usually the same as the ones under one which can run simulations but not always. If you are trying to run under a project account of which you are not a member your job will simply be denied.

To see the how many core years a project account has (how "large" it is) and to see if it was recently under or over used you can use the command:

    sshare -l -A account-name_cpu

You need to put the '_cpu' after the name of the account!!!


How to "read" the output of above commands

The column 'RawShares' shows you the size of the account in core years times 1000. You need to divide this number by 1000 to get the actual core years. The number of core years is the number of cores which can get used under the specific project account on average throughout the year. If the RawShares is 1 it means the account has not been used recently. This happens sometimes for the default accounts. Once you submitted a job the numbers will soon adjust.
The column 'LevelFS' shows you how much the account was recently used. A value above 1 means it is underused a number below 1 means it is overused.

The final priority of your job in the queue (wrt all other jobs) is a combination of the RawShares and the LeveFS => the higher the two numbers the higher the priority of your job.

I created myself a script to check all the accounts I am a member of. You are welcome to use it or take a copy of it. This is the script:

    ~winger/ovbin/qsp




  • Aucune étiquette
Écrire un commentaire...