Frequently asked questions

1. What are machine names?
   Machines are named with the pattern

   	    a SITE _ ROOM _ GROUP-CHARARACTER NUMBER
   

   which results in names such as

   	    a48-206-011  a48-206-m04
   

   The machines are grouped according to hardware characteristics which are described in the hardware groups table.
   
   
2. What machines are under PBS?
   All the machines are under PBS. If you require dedicated time on a specific machine or set of machines it can be removed from PBS. Please e-mail aces-support@acesgrid.org with your request. If you simply require interactive access then the PBS command "qsub -I" can be used (see queue help pages for more examples). To run on a particular machine the qsub option "-l" can be used with a command such as:

   	    $  qsub -I -l nodes=a48-206-018
   

   which would request an interactive session on the machine called a48-206-018.
   
   
3. How do I move files between machines?
   Within the ACES system cp can be used. To copy things onto and off ACES scp must be used. The pathnames to filesystems in ACES are described on the storage pages.
   
   
4. Is MatLAB available? And what about MatLAB licenses?
   Yes, both MatLAB R13 and R14 are available on the non-IA64 nodes. Please see the MatLAB pages for more details. The number of used and available MatLAB licenses can be seen by running:

   	    $  /usr/local/pkg/matlab/lmstat -a
   

   and the purchase of more licenses can be arranged -- please contact the ACES admins. We ask that users please refrain from unfairly hogging the number of available licenses. On the IA64 (Itanium) nodes and on the all the other nodes, the free Octave software is available. Octave can run many (but not all) MatLAB "*.m" scripts. For more Octave information, please see www.octave.org.
   
   
5. Is MatlabMPI available?
   Not yet...
   
   
6. Is Matlab*P available?
   Not yet...
   
   
7. What builds of netCDF are available?
   The latest from the NetCDF stable series (version 3.6.0beta) is currently compiled and installed with all compilers.
   
   
8. Why so many MPI variants? Which one do I use?
   Currently there are 4 main MPI implementations installed. Three of them are MPICH based, the other one is LAM. Each version was installed for its specific advantages and characteristics. Obviously if your code requires the use of MPE library calls or you plan to trace the execution of your code using the trace tools that come with MPICH, you have to use an MPICH-based implementation. Ditto for LAM extensions for debugging etc.
   • To avoid having to compile different binaries for different interconnects, use either MPICH-VMI or LAM. The binaries they generate work with either interconnect choosing the relevant code path dynamically.
   • To be able to use both interconnects together, utilizing Myrinet between Myrinet nodes, and Ethernet between the rest of the nodes, use MPICH-VMI. This enables the more optimal use of all the nodes in the clusters.
   • To use the Gigabit interconnect with the highest bandwidth and the lowest latency, use LAM.
   • To use the Myrinet interconnect with the highest bandwidth and the lowest latency, use MPICH-GM. The binary will only work with Myrinet.
   • To use the Ethernet interconnect with the lowest latency for an MPICH variant, use MPICH (which is off the main MPICH development tree).
   
   
   
9. Why do modules work interactively but sometimes fail in jobfiles?
   Depending upon how you authored your PBS jobfile, you may or may not invoke the shell scripts that initialize the modules information. To force initialization of modules, you can use this snippet for an sh or bash shell:

   	    # bash/sh
   	    if [ -f /etc/profile.d/modules.sh ]; then
                 . /etc/profile.d/modules.sh
   	    fi
   

   or this snippet for tcsh or csh shell:

   	    # tcsh/csh
   	    if ( -f /etc/profile.d/modules.csh ) then
   	      source /etc/profile.d/modules.csh
               endif
   

   within your PBS jobfile.
   
   
10. How can I use a program that has an X window display with PBS?
    One way to do this is to use PBS to "reserve" a set of nodes e.g.

             qsub -I -l nodes=1
    

    and then to use ssh to create a session with X connections forwarded on the nodes that PBS reserves. The command

             qstat -n
    

    can be used to show which specific nodes have been allocated to which PBS session.
    
    Please note that you may have to edit your SSH configuration (potentially, on both your client ***AND*** the ACES machines) to enable X forwarding over SSH. The lines you might want to add to your ${HOME}/.ssh/config file are:

             Host *
                     ForwardX11 yes
                     ForwardAgent yes
                     ForwardX11Trusted yes
    

    
    
    
11. Why is my shell script is giving me odd errors?
    This can be due to many things, good things to check are
    
    • Check the shell setting you are using at the start of the script the same as the shell syntax you are using.
    
    
12. When are the scheduled down-times? Or, why are some systems sometimes unavailable on Tuesday afternoons?
    In order to perform maintenance that requires some level of service disruption (eg. hardware upgrades, reboots for removal of hung processes, etc.), we have scheduled "routine street-sweeping" to happen on Tuesday afternoons (noon -- 5pm).