ARSC T3D Users' Newsletter 113, November 22, 1996

No news next week... Happy Thanksgiving!

MPP Queues Adjusted at ARSC

We have added 5-minute queues in the 8, 16, and 32 PE sizes. The goal is to help users in the development phase who, given the current load on ARSC's T3D, often find their debugging jobs queued up behind one or more 24-hour jobs.

We have also increased the priority of the 32 PE queues which have been blocked for days at a time by 8 and 16 PE jobs. The reasoning parallels that used when we gave the 64 PE queues the highest priority (see Newsletter #95 ). With a busy machine, the likelihood of a contiguous block of 64 or 32 PEs opening up when the 8 & 16 PE jobs get to jump in first, is pretty low.

We have also rewritten the queue ethic which attempts to define fair play amongst the T3D users. In this version, we give ARSC staff the right to enforce the policy. Don't worry, this is not a philosophical shift: we will deal with situations as usual, by contacting people by email and phone. The two purposes for this are to stress the importance of the rules and to cover ourselves in some extreme case. The only scenario I can think of in which we would actually hold a job is if someone submitted too many jobs (by mistake or through ignorance of the ethic) and then became unreachable. There hasn't been a single case when a T3D user didn't respond immediately when informed that they were violating the ethic.

The new ethic has three rules (see below). The first is simply a restatement of the old ethic. The second will ensure that a single user can't take over 75% of the T3D at a time by running jobs of all different sizes. The third is a cautious approach to the new 5 minute queues.

As always...

Feedback on our queue structure and management is always appreciated, and contributed greatly to the current set of changes. After a couple of weeks, these changes to the structure and ethic may be up for review.


Here is the official notice of the changes (on denali, "news t3dbatch"):
---------------------------------------------------------------------

      T3D Batch Queues
      ================
      The T3D batch queues were changed on Nov. 19, 1996. The current T3D
      queues (in priority order) are:

      Name of queue  Number  Maximum jobs   Time limit   When queue is
                     of PEs   per queue      on queue       enabled

      m_64pe_10m       64         1        10 minutes     always
      m_64pe_24h       64         1        24 hours       always
      m_32pe_5m        32         1         5 minutes     always
      m_32pe_24h       32         1        24 hours       always
      m_8pe_5m          8         1         5 minutes     always
      m_16pe_5m        16         1         5 minutes     always
      m_8pe_24h         8         2        24 hours       always
      m_16pe_24h       16         2        24 hours       always
      m_128pe_5m      128         1         5 minutes     Fri 6PM to Sun 4AM
                                                          and after downtimes
      m_128pe_8h      128         1         8 hours       Fri 6PM to Sun 4AM

      A request made to these queues will be run as soon as enough PEs are
      available to satisfy the request. T3D jobs run to completion without
      interruption.


      Specifying Queues
      =================
      When submitting a job to the T3D, you do not name one of the above
      queues.  Specify only the MPP pipe queue:

        #QSUB -q mpp

      Based on the number of PEs and time you request, NQS will route your
      job to the smallest/shortest possible queue.

      Queue Ethics
      ============
      The T3D is a popular, limited resource.  ARSC has found that T3D
      users are willing to work together to provide fair access to the
      queues.  Please follow these 'rules':

        1) Submit/execute no more than one job per 8 or 24 hour queue at a
           time.

        2) Do not submit/execute jobs in more than two different 24 hour
           queues at the same time.

        3) Submit/execute no more than two jobs per 5 or 10 minute queue
           at a time.

      As an example of the "queue ethic," if user "goodman" submitted a
      job to the m_64pe_24h queue, he/she would not submit another until
      the first had run to completion and thus made the queue available to
      other users. Meanwhile, "goodman" could have one job in the
      m_32pe_24h queue as well, without violating the ethic.  He/she could
      not have one in each of the m_32pe_24h and m_8pe_24h queues as well,
      however.

      Please contact 
User Services
 (consult@arsc.edu or 907-450-8602) if
      you feel that the queues are being misused, and we will work to
      resolve the situation.  ARSC may hold or delete jobs that are
      submitted in violation of the 'rules' outlined above.



      User's UDBSEE Limits
      ====================
      Most T3D users currently have a limit of 128 PEs for batch access
      and 8 PEs for interactive access.  Users can check their limits with
      the udbsee command:

            udbsee 
 grep jpelimit

      The output will indicate your limits in interactive(i) and
      batch(b).  For example:

               jpelimit[b]     :128:
               jpelimit[i]     :8:

      If your batch PE limit is too small to access the NQS queues and you
      would like to use them, please contact User Services.



      Tracking Your Jobs
      ==================
      Users can query the NQS batch system with the command:

            qstat -a

      to see what NQS T3D jobs are scheduled to run on the T3D. The
      utility mppview is available to see what jobs are currently running
      on the T3D.

                                             -- Tom Baring, Nov. 20, 1996

Quick-Tip Q & A


A:
 {{ Do programmers get carpal tunnel syndrome?  How bad is it?  Can you
      avoid it? }}

  Carpal tunnel syndrome is for everyone. My sense is that this is one
  of those injuries that provides plenty of east-to-ignore warning
  signs, and that can really mess up your life.  I personally know two
  people who have lost good careers in graphic design and programming
  due to CTS (not to mention losing hobbies, sports, etc.). Here is a
  Web site devoted to typing injuries:
     
http://www.cs.princeton.edu/~dwallach/tifaq/


  Thanks to the reader who sent this in:
  
  > Yes, it is a repetitive motion injury.  It is painful and I have lost
  > strength in my hands.  Doing simple things causes pain and aggravates
  > the injury, like holding a book to read it, holding a steering wheel
  > when driving, using a knife to cut something, writing with a pen or
  > pencil.
  > 
  > To try to avoid it, use ergonomic workstations (chairs with
  > adjustable arms, adjustable keyboard drawers with a soft
  > wrist-rest).  Be aware of your hand positions and try to keep them
  > neutral (not bent at the wrist).  Take frequent breaks (at least
  > every 45 minutes) to do something else briefly.
  > 
  > Once you have it, I've found physical therapy (ultra-sound, massage,
  > icing down the area, and strengthing and flexibility exercises) helps
  > a lot.  The wrist-braces you see everyone wear help some.  I've heard
  > that the software which allows a combination of voice commands and
  > keyboard usage helps some (but the technology still needs some
  > work).


Q:
 If you expect to create core or mppcore files, how can you arrange
   for them to be 
removed
 automatically, so they never use up quota
   in the current directory?

[ Answers, questions, and tips graciously accepted. ]


Current Editors:
Ed Kornkven ARSC HPC Specialist ph: 907-450-8669
Kate Hedstrom ARSC Oceanographic Specialist ph: 907-450-8678
Arctic Region Supercomputing Center
University of Alaska Fairbanks
PO Box 756020
Fairbanks AK 99775-6020
E-mail Subscriptions: Archives:
    Back issues of the ASCII e-mail edition of the ARSC T3D/T3E/HPC Users' Newsletter are available by request. Please contact the editors.
Back to Top