How to crunch Sztaki easier, tips and hints...


Advanced search

Message boards : SZTAKI Desktop Grid : How to crunch Sztaki easier, tips and hints...

AuthorMessage
Profile Nightbird
Forum moderator
Avatar
Send message
Joined: Jul 12 05
Posts: 920
Credit: 114,924
RAC: 0
Message 4378 - Posted 18 Oct 2006 23:15:28 UTC

    Last modified: 18 Oct 2006 23:27:42 UTC

    If that can help, feel free here to write your tips
    ____________

    Stick
    Send message
    Joined: Jun 12 06
    Posts: 193
    Credit: 66,244
    RAC: 0
    Message 4381 - Posted 19 Oct 2006 1:44:38 UTC - in response to Message 4378.

      If that can help, feel free here to write your tips


      This thread is a great idea! Unfortunately, I don\'t have a tip to add - but I do have a question whose answer maybe helpful to more people than just me.

      Is there a way to tell how many \"lines\" there are to process in a work unit?

      @ Nightbird: Since your intention here is to compile hints and tips (and not questions) please feel free to delete this post (hopefully, after someone provides the answer).
      ____________

      Odysseus
      Avatar
      Send message
      Joined: Feb 27 06
      Posts: 212
      Credit: 221,397
      RAC: 0
      Message 4384 - Posted 19 Oct 2006 6:41:47 UTC - in response to Message 4381.

        Last modified: 19 Oct 2006 6:52:32 UTC

        Is there a way to tell how many \"lines\" there are to process in a work unit?

        The most reliable is to open the input file with a text editor. The file will be named for that part of the WU’s name that comes between the underscores, preceded by “in.txt_”, and it will be found in the BOINC directory under /projects/szdg.lpds.sztaki.hu_szdg/. You needn’t have put “lines” in quotes, because that’s exactly what they are in the document. Each line comprises several groups of figures in square brackets, beginning with “pos #”, where # represents a number—these run consecutively down the file, so you don’t have to count them.

        The best tip I’ve seen recently was Nightbird’s strategy for dealing with excessive WU downloads—a very common problem here AFAICT—suspending all but a few that seem most likely to form a quorum.

        I’ve found the technique very helpful in the past few weeks while I’ve been dealing with the 47 WUs, initially estimated at just over an hour each, that my G5 Mac downloaded when I attached it to the project on Sept. 22. Although its RDF is now at 31.6, and many of its results are in an ‘orphan’ state, with the deadline less than a day away, it’s done much better than I expected at first: 20 completed (averaging about 24 h CPU time each), of those 9 validated, and of those 5 have already been cleared from the database. Just over a thousand credits were granted (about half what was claimed) and nearly 3500 pending—although I fear it will be a long time before some of those reach quorum. No errors on this host yet, and better still, no failed validations (touch wood). Moreover, I was able to keep BOINC out of EDF mode for a fair portion of the time, so my other projects haven’t suffered too badly.

        ____________

        Profile Nightbird
        Forum moderator
        Avatar
        Send message
        Joined: Jul 12 05
        Posts: 920
        Credit: 114,924
        RAC: 0
        Message 4428 - Posted 22 Oct 2006 20:31:40 UTC

          Last modified: 23 Oct 2006 0:38:41 UTC

          Your wu is sticky at xx.xx % ?
          Try that :
          suspend the wu, wait a bit (min, hours, days ??) then resume it.
          The % can be updated... and in the \'best of\' the % can jump at 100 %...

          link

          ____________

          Attalla
          Send message
          Joined: Sep 19 05
          Posts: 56
          Credit: 314,131
          RAC: 0
          Message 4443 - Posted 23 Oct 2006 17:43:33 UTC - in response to Message 4428.

            Your wu is sticky at xx.xx % ?
            Try that :
            suspend the wu, wait a bit (min, hours, days ??) then resume it.
            The % can be updated... and in the \'best of\' the % can jump at 100 %...

            link

            Szia!
            Ezzel csak egy probléma van: lsd. másik téma.
            Azóta 60%-ra lépett, de csak 2 sor készült el az 5-ből, tehát 40%-on van.
            Tehát nem oldja meg a problémát.
            Van esetleg más javaslat?
            Attis
            ____________

            Profile Nightbird
            Forum moderator
            Avatar
            Send message
            Joined: Jul 12 05
            Posts: 920
            Credit: 114,924
            RAC: 0
            Message 4446 - Posted 24 Oct 2006 3:17:06 UTC - in response to Message 4443.

              Your wu is sticky at xx.xx % ?
              Try that :
              suspend the wu, wait a bit (min, hours, days ??) then resume it.
              The % can be updated... and in the \'best of\' the % can jump at 100 %...

              link

              Szia!
              Ezzel csak egy probléma van: lsd. másik téma.
              Azóta 60%-ra lépett, de csak 2 sor készült el az 5-ből, tehát 40%-on van.
              Tehát nem oldja meg a problémát.
              Van esetleg más javaslat?
              Attis

              Hi Attalla,
              Do you know that my translator online is not excellent ?
              ____________

              Attalla
              Send message
              Joined: Sep 19 05
              Posts: 56
              Credit: 314,131
              RAC: 0
              Message 4449 - Posted 24 Oct 2006 7:02:57 UTC - in response to Message 4446.

                Your wu is sticky at xx.xx % ?
                Try that :
                suspend the wu, wait a bit (min, hours, days ??) then resume it.
                The % can be updated... and in the \'best of\' the % can jump at 100 %...

                link

                Szia!
                Ezzel csak egy probléma van: lsd. másik téma.
                Azóta 60%-ra lépett, de csak 2 sor készült el az 5-ből, tehát 40%-on van.
                Tehát nem oldja meg a problémát.
                Van esetleg más javaslat?
                Attis

                Hi Attalla,
                Do you know that my translator online is not excellent ?

                Szia!
                Nem baj. :)
                A probléma megoldódni. :D
                A számláló megint jól működni. :D
                Én is tanulni angol, te is tanulni magyar. :)
                fordító
                Attis

                Ui.: Bocs mindenkitől, de a szótárazók kedvéért muszáj. :D
                ____________

                Profile Ananas
                Send message
                Joined: Jul 12 05
                Posts: 222
                Credit: 665,833
                RAC: 0
                Message 4535 - Posted 29 Oct 2006 23:00:34 UTC

                  Last modified: 29 Oct 2006 23:01:04 UTC

                  In order to see that the WU is still doing \"something\" and not looping forever, it might be helpful to save a few of those \"dc_ckpt_##\" files (one at a time) from the slots directory of SZTAKI. If any of the ckpt files are identical, there might be an endless loop.

                  I have no idea how many steps such an endless loop can have until it repeats, in those samples I took I did not find an endless loop yet - but I just re-enabled SZTAKI, trying 2.01 on just one PC (currently my only SSE2 box) so my samples are sure not representative.

                  Philip Martin Kryder
                  Send message
                  Joined: Apr 3 06
                  Posts: 73
                  Credit: 10,801
                  RAC: 0
                  Message 4537 - Posted 30 Oct 2006 1:13:09 UTC - in response to Message 4535.

                    In order to see that the WU is still doing \"something\" and not looping forever, it might be helpful to save a few of those \"dc_ckpt_##\" files (one at a time) from the slots directory of SZTAKI. If any of the ckpt files are identical, there might be an endless loop.

                    I have no idea how many steps such an endless loop can have until it repeats, in those samples I took I did not find an endless loop yet - but I just re-enabled SZTAKI, trying 2.01 on just one PC (currently my only SSE2 box) so my samples are sure not representative.



                    More to the point -

                    Some might say that if the checkpoints (or any internal intermediate results) are identical, that situation should be detected by the software and the WU terminated....


                    ____________

                    Profile Ananas
                    Send message
                    Joined: Jul 12 05
                    Posts: 222
                    Credit: 665,833
                    RAC: 0
                    Message 4538 - Posted 30 Oct 2006 1:24:14 UTC

                      Last modified: 30 Oct 2006 1:26:26 UTC

                      Basically you are right and maybe SZTAKI even does it - but in some algorithms that isn\'t so easy as the repeated pattern can be quite long. The values I have seen are sometimes quite jumpy, so it might be compareable to the turing engine problem, where they try to create as long patterns as possible without running infinite - but without a chance to detect an asymptote or values heading towards infinity. Some Mandelbrot fractal areas show such a behaviour too, that\'s why you need to limit the number of iterations there.

                      Profile Ananas
                      Send message
                      Joined: Jul 12 05
                      Posts: 222
                      Credit: 665,833
                      RAC: 0
                      Message 4539 - Posted 30 Oct 2006 5:47:06 UTC

                        Last modified: 30 Oct 2006 6:08:32 UTC

                        Well, progress at 20% not moving at all, checkpoints restarting at 1 on restart of BOINC, runtime estimated way too low - I guess the project is still in the same state that caused me to stop crunching for it a few weeks ago - I shouldn\'t have returned I guess. *reset*

                        Profile Nightbird
                        Forum moderator
                        Avatar
                        Send message
                        Joined: Jul 12 05
                        Posts: 920
                        Credit: 114,924
                        RAC: 0
                        Message 4546 - Posted 30 Oct 2006 23:25:59 UTC

                          Last modified: 31 Oct 2006 0:15:21 UTC

                          For people who have (still) wus running with the application 2.00 and IF your cpu supports SSE2 instructions, you can try (at your own risk) the test version 2.02 :

                          1. Download the application 2.02 from here
                          2. Exit BOINC
                          3. Delete the existing program \"search_2.00_windows_intelx86.exe\" from your \"C:\\Program Files\\BOINC\\projects\\szdg.lpds.sztaki.hu_szdg\" folder
                          or rename it \"search_2.00_windows_intelx86.exe.old\"
                          4. Copy the new program \"search_2.02_windows_intelx86.exe\" to that same folder
                          5. Rename the new program to back to \"search_2.00_windows_intelx86.exe\"
                          6. Restart BOINC

                          That can help with some long wus.

                          (thanks to Stick for the how-to)
                          ____________

                          Philip Martin Kryder
                          Send message
                          Joined: Apr 3 06
                          Posts: 73
                          Credit: 10,801
                          RAC: 0
                          Message 4548 - Posted 31 Oct 2006 5:36:15 UTC - in response to Message 4546.

                            Last modified: 31 Oct 2006 5:39:09 UTC

                            For people who have (still) wus running with the application 2.00 and IF your cpu supports SSE2 instructions, you can try (at your own risk) the test version 2.02 :

                            ....


                            Just to be clear, are you only recommending 2.02 for folks still on 2.00?

                            If we are already on 2.01
                            search_2.01_windows_intelx86.exe
                            (or is that 2.03???) are we at the most current production version?

                            It might be nice to just post the current \"production\" version
                            and
                            the current \"run at your own risk\" version....

                            Though there is a part of me that feels this whole project is \"run at your own risk\" - which is ok, since so much of life is like that...

                            ____________

                            Profile Nightbird
                            Forum moderator
                            Avatar
                            Send message
                            Joined: Jul 12 05
                            Posts: 920
                            Credit: 114,924
                            RAC: 0
                            Message 4554 - Posted 31 Oct 2006 21:14:21 UTC - in response to Message 4548.

                              Last modified: 31 Oct 2006 21:18:49 UTC

                              For people who have (still) wus running with the application 2.00 and IF your cpu supports SSE2 instructions, you can try (at your own risk) the test version 2.02 :

                              ....


                              Just to be clear, are you only recommending 2.02 for folks still on 2.00?

                              If we are already on 2.01
                              search_2.01_windows_intelx86.exe
                              (or is that 2.03???) are we at the most current production version?

                              It might be nice to just post the current \"production\" version
                              and
                              the current \"run at your own risk\" version....

                              Though there is a part of me that feels this whole project is \"run at your own risk\" - which is ok, since so much of life is like that...

                              The 2.02 for folks still on 2.00.

                              The 2.00 is an official version.
                              The 2.01 is an official version but it crashes on some cpus (without SSE2 instructions).
                              The 2.02 (and the 2.03) are (for me) test versions or \"at your own risk\" but we can hope that it runs (without too many problems (?)) on some cpus (with SSE2 instructions). Something is sure : the application is faster.

                              So maybe we can try to use a 2.02 or 2.03 renaming them 2.00.
                              Aim : trying to accelerate some \"recalcitrant\" wus like 1a9d6387 and avoid the \"Maximum CPU time exceeded\".

                              I had suspended on my Athlon64 X2 4600+ (running the 2.00) wus 1a9d6387.
                              I tried the 2.02.
                              I succeeded with 2 wus 1a9d6387 but after 2 reboots, i lost 9 wus (\"computation couldn\'t start\"). I don\'t know why.
                              So i\'m careful with the idea.
                              Perhaps you will succeed, perhaps you will fail.
                              \"at your own risk\"

                              ____________

                              Philip Martin Kryder
                              Send message
                              Joined: Apr 3 06
                              Posts: 73
                              Credit: 10,801
                              RAC: 0
                              Message 4555 - Posted 1 Nov 2006 4:55:35 UTC - in response to Message 4554.

                                For people who have (still) wus running with the application 2.00 and IF your cpu supports SSE2 instructions, you can try (at your own risk) the test version 2.02 :

                                ....


                                Just to be clear, are you only recommending 2.02 for folks still on 2.00?

                                If we are already on 2.01
                                search_2.01_windows_intelx86.exe
                                (or is that 2.03???) are we at the most current production version?

                                It might be nice to just post the current \"production\" version
                                and
                                the current \"run at your own risk\" version....

                                Though there is a part of me that feels this whole project is \"run at your own risk\" - which is ok, since so much of life is like that...

                                The 2.02 for folks still on 2.00.

                                The 2.00 is an official version.
                                The 2.01 is an official version but it crashes on some cpus (without SSE2 instructions).
                                The 2.02 (and the 2.03) are (for me) test versions or \"at your own risk\" but we can hope that it runs (without too many problems (?)) on some cpus (with SSE2 instructions). Something is sure : the application is faster.

                                So maybe we can try to use a 2.02 or 2.03 renaming them 2.00.
                                Aim : trying to accelerate some \"recalcitrant\" wus like 1a9d6387 and avoid the \"Maximum CPU time exceeded\".

                                I had suspended on my Athlon64 X2 4600+ (running the 2.00) wus 1a9d6387.
                                I tried the 2.02.
                                I succeeded with 2 wus 1a9d6387 but after 2 reboots, i lost 9 wus (\"computation couldn\'t start\"). I don\'t know why.
                                So i\'m careful with the idea.
                                Perhaps you will succeed, perhaps you will fail.
                                \"at your own risk\"



                                why the renaming rather than just release a new version with the automatic download?




                                ____________

                                Profile Nightbird
                                Forum moderator
                                Avatar
                                Send message
                                Joined: Jul 12 05
                                Posts: 920
                                Credit: 114,924
                                RAC: 0
                                Message 4583 - Posted 5 Nov 2006 0:58:49 UTC

                                  It\'s no new version presently so i tried an idea and not bad, i succeeded in finishing 4 wus 1a9d6387-.
                                  ____________

                                  Profile Nightbird
                                  Forum moderator
                                  Avatar
                                  Send message
                                  Joined: Jul 12 05
                                  Posts: 920
                                  Credit: 114,924
                                  RAC: 0
                                  Message 4591 - Posted 5 Nov 2006 22:07:43 UTC - in response to Message 4449.

                                    Last modified: 5 Nov 2006 22:12:07 UTC


                                    Szia!
                                    Nem baj. :)
                                    A probléma megoldódni. :D
                                    A számláló megint jól mu\"ködni. :D
                                    Én is tanulni angol, te is tanulni magyar. :)
                                    fordító
                                    Attis

                                    Ui.: Bocs mindenkito\"l, de a szótárazók kedvéért muszáj. :D

                                    Attis, thanks for the link (didn\'t know this site) but now i have a question :
                                    how do i do from magyar to angol ?

                                    ____________

                                    Attalla
                                    Send message
                                    Joined: Sep 19 05
                                    Posts: 56
                                    Credit: 314,131
                                    RAC: 0
                                    Message 4597 - Posted 6 Nov 2006 22:23:02 UTC - in response to Message 4591.


                                      Szia!
                                      Nem baj. :)
                                      A probléma megoldódni. :D
                                      A számláló megint jól mu\"ködni. :D
                                      Én is tanulni angol, te is tanulni magyar. :)
                                      fordító
                                      Attis

                                      Ui.: Bocs mindenkito\"l, de a szótárazók kedvéért muszáj. :D

                                      Attis, thanks for the link (didn\'t know this site) but now i have a question :
                                      how do i do from magyar to angol ?

                                      Szia!
                                      Bocsi! Igazad van. visszafelé nem működik. :)
                                      Attis
                                      ____________

                                      Philip Martin Kryder
                                      Send message
                                      Joined: Apr 3 06
                                      Posts: 73
                                      Credit: 10,801
                                      RAC: 0
                                      Message 4600 - Posted 7 Nov 2006 4:44:01 UTC - in response to Message 4538.

                                        Basically you are right and maybe SZTAKI even does it - but in some algorithms that isn\'t so easy as the repeated pattern can be quite long. The values I have seen are sometimes quite jumpy, so it might be compareable to the turing engine problem, where they try to create as long patterns as possible without running infinite - but without a chance to detect an asymptote or values heading towards infinity. Some Mandelbrot fractal areas show such a behaviour too, that\'s why you need to limit the number of iterations there.


                                        do you mean \"long\" in terms of number of iterations between repetitions?

                                        Or do you mean long in terms of the amount of data needed to represent the intermediate result?

                                        One technique is to keep a hash total and only compare the full amount of data when the hash totals match.
                                        ____________

                                        Philip Martin Kryder
                                        Send message
                                        Joined: Apr 3 06
                                        Posts: 73
                                        Credit: 10,801
                                        RAC: 0
                                        Message 4601 - Posted 7 Nov 2006 4:44:48 UTC - in response to Message 4583.

                                          It\'s no new version presently so i tried an idea and not bad, i succeeded in finishing 4 wus 1a9d6387-.


                                          were you running 2.02 or
                                          2.03 renamed as 2.01?

                                          thanks!
                                          Phil
                                          ____________

                                          Profile Nightbird
                                          Forum moderator
                                          Avatar
                                          Send message
                                          Joined: Jul 12 05
                                          Posts: 920
                                          Credit: 114,924
                                          RAC: 0
                                          Message 4608 - Posted 7 Nov 2006 23:58:54 UTC - in response to Message 4601.

                                            It\'s no new version presently so i tried an idea and not bad, i succeeded in finishing 4 wus 1a9d6387-.


                                            were you running 2.02 or
                                            2.03 renamed as 2.01?

                                            thanks!
                                            Phil

                                            I\'m running an Athlon64 X2 but with the 2.00 it\'s a snail.
                                            So i installed the 2.02 renamed 2.00 to \"help\" some wus 1a9d6387.




                                            ____________

                                            Philip Martin Kryder
                                            Send message
                                            Joined: Apr 3 06
                                            Posts: 73
                                            Credit: 10,801
                                            RAC: 0
                                            Message 4610 - Posted 8 Nov 2006 3:21:43 UTC - in response to Message 4608.

                                              It\'s no new version presently so i tried an idea and not bad, i succeeded in finishing 4 wus 1a9d6387-.


                                              were you running 2.02 or
                                              2.03 renamed as 2.01?

                                              thanks!
                                              Phil

                                              I\'m running an Athlon64 X2 but with the 2.00 it\'s a snail.
                                              So i installed the 2.02 renamed 2.00 to \"help\" some wus 1a9d6387.





                                              so, if we are already running 2.01 is there any reason to do this?

                                              ____________

                                              Profile Nightbird
                                              Forum moderator
                                              Avatar
                                              Send message
                                              Joined: Jul 12 05
                                              Posts: 920
                                              Credit: 114,924
                                              RAC: 0
                                              Message 4650 - Posted 12 Nov 2006 17:19:55 UTC - in response to Message 4610.

                                                Last modified: 12 Nov 2006 19:02:27 UTC

                                                It\'s no new version presently so i tried an idea and not bad, i succeeded in finishing 4 wus 1a9d6387-.


                                                were you running 2.02 or
                                                2.03 renamed as 2.01?

                                                thanks!
                                                Phil

                                                I\'m running an Athlon64 X2 but with the 2.00 it\'s a snail.
                                                So i installed the 2.02 renamed 2.00 to \"help\" some wus 1a9d6387.





                                                so, if we are already running 2.01 is there any reason to do this?

                                                This A64 X2 was running with the 2.00. Without no new wus downloaded, the new application 2.01 wasn\'t installed on this machine.

                                                ____________

                                                Profile Nightbird
                                                Forum moderator
                                                Avatar
                                                Send message
                                                Joined: Jul 12 05
                                                Posts: 920
                                                Credit: 114,924
                                                RAC: 0
                                                Message 4651 - Posted 12 Nov 2006 17:24:54 UTC

                                                  Last modified: 12 Nov 2006 18:27:58 UTC

                                                  If possible, don\'t use a too large work cache size (Connect to network about every xx days).
                                                  (I\'ve seen a machine with ~ 1600 wus downloaded between 1 and 8 nov.)





                                                  ____________

                                                  Profile Nightbird
                                                  Forum moderator
                                                  Avatar
                                                  Send message
                                                  Joined: Jul 12 05
                                                  Posts: 920
                                                  Credit: 114,924
                                                  RAC: 0
                                                  Message 4696 - Posted 19 Nov 2006 17:36:38 UTC

                                                    Last modified: 19 Nov 2006 21:19:39 UTC

                                                    Since i had 3 (difficult ?) wus (using the application 2.01) running on my A64 X24600+.
                                                    - 1 wu 1a9d6387 wuid=16198 : sticky at 80 % since many days now
                                                    - 2 wus 21b342bd wuid=16438 wuid=16448 running since some days but with a completion time = cpu time done and a % done = 0
                                                    I closed Boinc and rebooted my machine.
                                                    And guess ?
                                                    The 3 wus finished immediately and the % done = 100 %.
                                                    wu 1a9d6387 : 1 day 2h 27min 34sec
                                                    wu 21b342bd : 2 days 4h 07min 14sec
                                                    wu 21b342bd : 6h 36min 39 sec

                                                    Well my A64 X2 is a snail. ;)

                                                    Edit :
                                                    Same problem with my A64 X4400+ (using the new 2.02) : the % done = 0 for 5 wus
                                                    Shut down Boinc and rebooted.
                                                    Immediately, the wuid=13878 finished.
                                                    About the 4 others, % done and completion time were updated (3 wus at 20 % and 1 wu at 18 %).

                                                    ____________

                                                    larry1186
                                                    Send message
                                                    Joined: Sep 25 06
                                                    Posts: 37
                                                    Credit: 18,502
                                                    RAC: 0
                                                    Message 5556 - Posted 25 Jan 2007 15:45:04 UTC

                                                      Here are some tips that a user can do to give their WUs the best chance for valid results:


                                                      • Set general prefs Leave applications in memory while preempted? to YES.
                                                      • Do not exit BOINC if possible. That means running your hosts 24/7 and, sorry, no more daily reboots.
                                                      • If you feel that the % complete is stuck - don\'t worry, just look at CPU time for reassurance that the WU is still going. The % complete is representative of the # of lines completed (e.g. a 5-line WU will jump from 0% to 20% when the first line is complete, then to 40% upon completion of the second line, and so on. A 1-line WU will only jump to 100% complete when it is actually done. Keep in mind that each line will take a totally different amount of time to complete)
                                                      • Ignore the \"To completion\" time, it\'s irrelevant due to the nature of the algorithm used in the application.
                                                      • If you are still nervous about a WU even if the CPU time is still counting, abort it. If you stop BOINC and restart to get the WU going again, the result will be no good and will not validate and you won\'t get credit for it anyway. Although I do encourage you to stay with it (it may take 100+ hours), but if you aren\'t comfortable with it, then don\'t bother. Worst case for staying with it = Max CPU time error.
                                                      • Be patient.
                                                      • Expect to lose some credit due to WUs getting too many success/total/error results. (the too many success or total results happens when hosts produce successful yet invalid results, usually {but not always} by stopping/restarting BOINC. By aborting it produces an error result.)



                                                      Hope this helps some people that are feeling a bit uneasy with this project. Happy Crunching!
                                                      Please note, these are the views of the author of this post and are in no way representative of the community at SZTAKI
                                                      ____________
                                                      Don't get distracted by shiny objects.

                                                      robert.mouris
                                                      Send message
                                                      Joined: Nov 3 05
                                                      Posts: 129
                                                      Credit: 4,124,189
                                                      RAC: 0
                                                      Message 5560 - Posted 25 Jan 2007 18:43:54 UTC - in response to Message 5556.

                                                        This is a wonderful description of what we have experienced and learned during the last weeks. Could an administrator insert a link on the homepage and the FAQ, thus making it official and visible? It might rebuff a few potential crunchers, but there is no point in making them produce unusable results and getting unhappy. The project will win by having less, but informed and happy crunchers.

                                                        I would just say one thing differently, as we must all reboot our system from time to time. Each time I have downloaded (too) many WUs, I set the project to \"no new tasks\", and a few hours or days before I want to stop, I suspend all the waiting WUs before they start crunching. Rebooting the system does not hurt WUs that haven\'t started yet. The sentence \"That means running your hosts 24/7 and, sorry, no more daily reboots.\" is a bit frightening.


                                                        ____________

                                                        Ironheart
                                                        Send message
                                                        Joined: Jan 16 07
                                                        Posts: 20
                                                        Credit: 4,366
                                                        RAC: 0
                                                        Message 6233 - Posted 10 May 2007 0:59:47 UTC - in response to Message 5556.

                                                          If only this were posted on the home page of the project I might have not wasted
                                                          242,042.58 CPU time (sec) for a granted credit of 0.00 on Computer 267105 for Result ID 809118 for WU ID 181264 I mean who would rationally crunch so long for a result that will be no good and will not validate? ...

                                                          But this info is not on the home page so ... oh well ...



                                                          Here are some tips that a user can do to give their WUs the best chance for valid results:

                                                          • Set general prefs Leave applications in memory while preempted? to YES.
                                                          • Do not exit BOINC if possible. That means running your hosts 24/7 and, sorry, no more daily reboots.
                                                          • If you feel that the % complete is stuck - don\'t worry, just look at CPU time for reassurance that the WU is still going. The % complete is representative of the # of lines completed (e.g. a 5-line WU will jump from 0% to 20% when the first line is complete, then to 40% upon completion of the second line, and so on. A 1-line WU will only jump to 100% complete when it is actually done. Keep in mind that each line will take a totally different amount of time to complete)
                                                          • Ignore the \"To completion\" time, it\'s irrelevant due to the nature of the algorithm used in the application.
                                                          • If you are still nervous about a WU even if the CPU time is still counting, abort it. If you stop BOINC and restart to get the WU going again, the result will be no good and will not validate and you won\'t get credit for it anyway. Although I do encourage you to stay with it (it may take 100+ hours), but if you aren\'t comfortable with it, then don\'t bother. Worst case for staying with it = Max CPU time error.
                                                          • Be patient.
                                                          • Expect to lose some credit due to WUs getting too many success/total/error results. (the too many success or total results happens when hosts produce successful yet invalid results, usually {but not always} by stopping/restarting BOINC. By aborting it produces an error result.)



                                                          Hope this helps some people that are feeling a bit uneasy with this project. Happy Crunching!
                                                          Please note, these are the views of the author of this post and are in no way representative of the community at SZTAKI


                                                          Post to thread

                                                          Message boards : SZTAKI Desktop Grid : How to crunch Sztaki easier, tips and hints...


                                                          Home | My Account | Message Boards


                                                          Copyright © 2017 SZTAKI Desktop Grid