Patience Please. my .02


Advanced search

Message boards : SZTAKI Desktop Grid : Patience Please. my .02

AuthorMessage
Profile Shane Phillips
Avatar
Send message
Joined: Jan 17 06
Posts: 19
Credit: 25,307
RAC: 0
Message 3742 - Posted 7 Aug 2006 1:29:58 UTC

    Last modified: 7 Aug 2006 1:30:15 UTC

    What a shock! I read the board and I finally got the issue. While there have been some issues with WU deadline and Linux WU, the real issue here is that this phase of the SZTAKI project is VERY different.

    We went from <8hr/WU to >80-100hrs/WU. The expectation of finishing a WU and getting credit within the day is gone. I just turned in a 90 hour WU. http://szdg.lpds.sztaki.hu/szdg/workunit.php?wuid=39650
    Since I am the 1 of 3 quorum, I can probably wait several weeks to get my 690 pending credits since it has to go to 2 others and they have to run it with other projects.

    I find this similar to Climate Prediction vs Einstein for time (Yes, I realize CPDN gives trickle credits along the way).

    Therefore, everyone please stick with it. Otherwise it will take even long for quorums of results to get done.

    I have 5 laptops running BOINC. I have bascially put 2 full time on SZTAKI. Therefore, my 90 hr WU really to 90 hours (3.77 days).

    My goal is to participate (>10,000 credits/project) in as many boinc projects as possible. So far so good. http://boinc.mundayweb.com/one/stats.php?userID=3580

    It may take awhile, but I am sticking wiht SZTAKI.

    Happy processing.

    I will be patient with SZTAKI.
    ____________

    Don Boyd
    Send message
    Joined: Feb 18 06
    Posts: 3
    Credit: 16,623
    RAC: 0
    Message 3784 - Posted 10 Aug 2006 22:45:30 UTC

      I don\'t mind the long WUs...as long as we are given enough time to complete them. I just had a WU that was going to take 30000 hours to complete. Even running 24 hours a day this is over 100 days. When I aborted it it was at .2% and had only ten days to complete. If they are going to send out these Super WUs they need to set an adequate deadline. If you see one of thesxe Super Wus do not even bother with it abort the thing and post why you have done so...maybe the people who run this place will get the message.
      ____________

      Odysseus
      Avatar
      Send message
      Joined: Feb 27 06
      Posts: 212
      Credit: 221,397
      RAC: 0
      Message 3785 - Posted 11 Aug 2006 0:35:29 UTC - in response to Message 3784.

        I don\'t mind the long WUs...as long as we are given enough time to complete them. I just had a WU that was going to take 30000 hours to complete. Even running 24 hours a day this is over 100 days. When I aborted it it was at .2% and had only ten days to complete. If they are going to send out these Super WUs they need to set an adequate deadline. If you see one of thesxe Super Wus do not even bother with it abort the thing and post why you have done so...maybe the people who run this place will get the message.

        I\'ve had to do that twice now, and reported it both times; I\'ve got two more WUs that have made a little progress (after over a week in \"panic mode\") but probably not enough, so will also have to be killed if they don\'t show signs of improvement in the next few days.

        Meanwhile, I have also been recommending patience, and was rewarded a couple of days ago by getting a \"Maximum CPU time exceeded error\" about 90% of the way through a WU after 273 hours of crunching. :(
        ____________

        WimTea
        Send message
        Joined: Mar 27 06
        Posts: 17
        Credit: 2,773
        RAC: 0
        Message 3789 - Posted 11 Aug 2006 6:57:25 UTC

          Last modified: 11 Aug 2006 7:02:59 UTC

          Found a little tweak to prevent Max_CPU_time_exceeded errors for unexpectedly long running WUs like some Sztake ones:
          1- Stop BOINC (service or regular background process)
          2- Go to BOINC main dir
          3- Edit client_state.xml with a plain text editor like notepad
          4- Locate rsc_fpops_bound. It\'s 3 lines below any Sztaki WU name.
          5- Shift the decimal point one place to the right (increase the value by a factor 10). Be sure you get this right!
          6- Save and exit client_state.xml
          7- Restart BOINC

          You\'ll have to repeat steps 4 and 5 for every Sztaki WU.

          Beware of the possibility that a WU becomes caught in a loop, as BOINC will then spend a LOT of extra CPU cycles on it before aborting it.
          ____________

          WimTea
          Send message
          Joined: Mar 27 06
          Posts: 17
          Credit: 2,773
          RAC: 0
          Message 3790 - Posted 11 Aug 2006 7:01:55 UTC - in response to Message 3789.

            Last modified: 11 Aug 2006 7:02:31 UTC

            <deleted>
            ____________

            Profile XJR-Maniac
            Avatar
            Send message
            Joined: Feb 9 06
            Posts: 27
            Credit: 225,973
            RAC: 0
            Message 3809 - Posted 14 Aug 2006 22:34:59 UTC - in response to Message 3789.

              Found a little tweak to prevent Max_CPU_time_exceeded errors for unexpectedly long running WUs like some Sztake ones:
              1- Stop BOINC (service or regular background process)
              2- Go to BOINC main dir
              3- Edit client_state.xml with a plain text editor like notepad
              4- Locate rsc_fpops_bound. It\'s 3 lines below any Sztaki WU name.
              5- Shift the decimal point one place to the right (increase the value by a factor 10). Be sure you get this right!
              6- Save and exit client_state.xml
              7- Restart BOINC

              You\'ll have to repeat steps 4 and 5 for every Sztaki WU.

              Beware of the possibility that a WU becomes caught in a loop, as BOINC will then spend a LOT of extra CPU cycles on it before aborting it.


              Good hint, tried it with the remaining WUs.

              Is there something to do when a WU is caught in a loop? Got some WUs over 200 hours and still under 10%!
              ____________

              robert.mouris
              Send message
              Joined: Nov 3 05
              Posts: 129
              Credit: 4,124,189
              RAC: 0
              Message 3840 - Posted 19 Aug 2006 3:53:03 UTC - in response to Message 3789.

                Last modified: 19 Aug 2006 3:53:59 UTC

                Found a little tweak to prevent Max_CPU_time_exceeded errors for unexpectedly long running WUs like some Sztake ones:
                1- Stop BOINC (service or regular background process)
                2- Go to BOINC main dir
                3- Edit client_state.xml with a plain text editor like notepad
                4- Locate rsc_fpops_bound. It\'s 3 lines below any Sztaki WU name.
                5- Shift the decimal point one place to the right (increase the value by a factor 10). Be sure you get this right!
                6- Save and exit client_state.xml
                7- Restart BOINC


                That\'s what I did. And this is what I got now after 201 hours:

                19/08/2006 05:02:25|SZTAKI Desktop Grid|Aborting result ab094285-a60b-418b-a548-bec0c0d6271d_cdce2fa8-3ff9-4e83-8abd-f6a89e75e29f_1571_2: exceeded CPU time limit 725625.000000
                19/08/2006 05:02:25|SZTAKI Desktop Grid|Unrecoverable error for result ab094285-a60b-418b-a548-bec0c0d6271d_cdce2fa8-3ff9-4e83-8abd-f6a89e75e29f_1571_2 (Maximum CPU time exceeded)

                :-(

                ____________

                Post to thread

                Message boards : SZTAKI Desktop Grid : Patience Please. my .02


                Home | My Account | Message Boards


                Copyright © 2017 SZTAKI Desktop Grid