Information for new users:

(This file is a copy of the new users template message, as of September 2002
With the telnet info removed.)

The MACI facility consists of five SGI Origin machines with a current
total of 236 processors:

     aurora    44 (MIPS R1000 195 MHz) processors with 11.75G of RAM 
     borealis  64 (MIPS R12000 400 MHz) processors with 16G of RAM
     australis 64 (MIPS R12000 400 MHz) processors with 32G of RAM
     helios    32 (MIPS R14000 500 MHz) processors with 16G of RAM
     *corona   32 (MIPS R14000 500 MHz) processors with 16G of RAM
     (*At the present time usage of corona is reserved for exclusive
       use of Dr. Rankin and his group.)


Acknowledgement Request from MACI:
---------------------------------
     The MACI facility is funded by the CFI, IIPP, UofA (CNS), C3,
AND local researchers.  It is imperative that MACI be kept informed
about, and acknowledged in, any published papers, conference 
presentations, and theses which benefit from the usage of the MACI
facilities. Our continued funding is dependent upon your compliance
to this request. Reprints and e-mail pointing out the above would be
appreciated.

     Some researchers have requested a sample acknowledgement for
inclusion in publications containing results generated using MACI 
facilities. We suggest the following (and please feel free to modify
it to suit your needs):
 
  Our work made use of the infrastructure and resources of MACI
  (Multimedia Advanced Computational Infrastructure), funded in
  part by the CFI (Canada Foundation for Innovation), ISRIP 
  (Alberta Innovation and Science Research Investment Program),
  and the Universities of Alberta and Calgary.


Conditions of Use:
-----------------
     This is a University of Alberta computing account and we would like
to direct you to a web publication you should read:
            http://www.ualberta.ca/CNS/policy/cns_conditions_of_use.html

We expect all new users to send us an email confirming that you have 
read this document and that you agree to comply with the CNS conditions
of use.


Changing Your Password and Forwarding Mail:
------------------------------------------
If you need to change your password please use the following url:
               https://ldap.srv.ualberta.ca/PROFILE

Enter your AFS/CNS id and AFS/CNS password, once authenticated please
follow online instructions to change your password and/or redirect 
your email. 


Disk Space:
----------
     Working disk storage space is provided in /scratch located on
a distributed high performance, CXFS, filesystem common to all the
MACI SGI machines. We have already created for you a directory named 
/scratch/youruserid. You simply 'cd /scratch/youruserid' and you are 
ready to go. Your $HOME and AFS disk space is limited to 5MB because
we do not expect people to work in that area. Please remember to 
remove files when they are no longer needed. Also, note that /scratch
is considered to be temporary space, and as such, please insure that
all of your important data is archived elsewhere.  



Getting Started:
---------------
     Interactive access, via ssh, is allowed on only two processors
on Aurora. The other processors on Aurora, as well as all processors on
Borealis, Australis, Corona and Helios, are run in a batch only 
environment using the Portable Batch System (PBS) for job distribution.
(See the following sections for more details.)

     For more information on ssh, see our local web page:
          http://www.ualberta.ca/CNS/RESEARCH/SSH
                     
     For details on the usage rules, priority calculation and how the
queues are sorted, please read our local web page: 
          http://www.ualberta.ca/CNS/RESEARCH/MACI/Scheduling.Pol.html

     Small  interactive jobs are allowed on Aurora, but are limited to
2 CPUs and less than 60 minutes of CPU time. Any larger jobs must be 
submitted using PBS. (We will send out reminders to anyone that we see 
not using the batch software.) Using PBS is not that difficult, however
you will need to write submission scripts. 

     To learn how to submit batch request and how to view job status 
please read our local users guide on: Portable Batch System (PBS)


Environment:
-----------
     To assist you with setting up the environment on this machine, we
have included .login and .cshrc in your home directory as well as a
soft link to your AFS space. 


E-mail Users Group:
------------------
     As a member of our High Performance Computer users group 
(oexternal-"Origin-external users") you will be receiving occasional
mail about updates and scheduled down times for this machine.


Accounting Information:
----------------------

Information about fair share priority of your group can be obtained
using the following command:

    aurora:> priorities

The priority of a group is calculated by dividing the group's shares
by the group's recent usage.

Again, more information about Policies is available at:
       http://www.ualberta.ca/CNS/RESEARCH/MACI/Scheduling.Pol.html
 

More information:
----------------
     More information, documentation, available software, on-line 
courses, etc., can be found on our web site:
               http://www.ualberta.ca/CNS/RESEARCH/MACI/

Any questions and/or concerns regarding aurora should be mailed to:
                     research.support@ualberta.ca

------------------------------------------------------------------------

 
JOB SUBMISSION FOR MACI SGI MACHINES USING PBS:
----------------------------------------------  

     To submit your long running jobs please use the Portable Batch
System (PBS) software. PBS is used to help everyone receive their fair
share of cpu time on the Origin Complex. For more information on PBS, 
please visit our web pages at:
                http://www.ualberta.ca/CNS/RESEARCH/PBS/

Small interactive jobs will be allowed on Aurora, but are limited to 2
CPUs and less than 60 minutes of CPU time. Larger jobs must be submitted
using PBS. We will send out reminders to anyone that does not comply.

Using PBS is straightforward. However, you will need to write submission
scripts (see our PBS web page for full instructions and examples). Once
the script is written, the basic commands are: 'qsub' to submit
programs; 'qstat' to check what is running; and 'qdel' to cancel 
submitted jobs.

BASIC PBS COMMANDS:
------------------
This summary of the basic commands is also available on our PBS web
site, and please feel free to contact us by email. (Also, use the 
'man' command on aurora to learn more about other PBS commands.)

  JOB SUBMISSION: qsub
  -------------------- 
  qsub [ option ] [ script-file ]

  Some eligible formats are:
   aurora:> qsub script_name
   aurora:> qsub -l ncpus=8 -l mem=2gb -l walltime=1:10:00 p_name
   aurora:> qsub -l ncpus=8 script_name

  In a submission script, PBS options are preceded by #PBS
  Here is an example of a typical line of a PBS script:
          #PBS -l ncpus=8
  Following #PBS is the option field and it is usually in the format:
          -flag [ param=value[,param=value,...] ]
  (without spaces after the comma).
  For example:
          -l ncpus=8,mem=2gb


  JOB CONTROL: qstat
  ------------------
  Information about jobs that have been submitted can be obtained
  with the qstat command. Examples are:
    aurora:> qstat                -- shows only basic information
    aurora:> qstat -f             -- shows full information
    aurora:> qstat -f -Q          -- shows full information about
                                        all the available queues
    aurora:> qstat -f -Q aurora   -- shows full information about
                                        the queue 'aurora'
  JOB CONTROL: qdel
  -----------------
  A job may either be stopped/killed by the user or by the 
  pbs-server. If the user decides to kill his/her job he can
  use the qdel command.
  For example:
    aurora:> qdel 1234
  where '1234' is the job ID, obtained from the command qstat.

  Alternatively, the job will be stopped when its execution time
  reaches either the time limits set in the submission script (cput
  or walltime) or the time limits of the queue on which the job runs.
  Currently, queues on aurora, australis and helios have a walltime
  limit of 24 hours; on borealis the limit is 12 hours.

  If you estimate that a particular job will run longer than the
  24/12 hours, then you will need to incorporate checkpointing and 
  signal handling into your program.

  Information on checkpointing and signal handling can be found on
  our PBS web page:
              http://www.ualberta.ca/CNS/RESEARCH/PBS/


SUBMITTING JOBS TO THE BATCH ONLY MACHINES:
------------------------------------------
BOREALIS, AUSTRALIS and HELIOS: 
-------------------------------
(The following information, along with sample scripts, is available 
 on our PBS web page.)

     In accordance with the MACI Executive decision on scheduling
polices, Borealis, Australis and Helios operates in a batch only 
environment used exclusively for parallel jobs. Users are not permitted
to directly login to these hosts.

Currently a PBS script needs to specify which machine you want the job
to run on, so you will need to include one of the following lines in
your script:

        #PBS -l host=aurora
    or
        #PBS -l host=borealis
    or
        #PBS -l host=australis
    or
        #PBS -l host=helios
   (NOTE: The default is aurora.)

The PBS script should also specify a mintime. Here is an example
of the mintime set to 1 hour:
        #PBS -l mintime=00:01:00

Again, if you have any problems with MACI's SGI High Performance
Computing complex; with how to write PBS scripts; with how to run
parallel jobs; or if you have any comments or concerns about the
policy, please email us at: research.support@ualberta.ca