Advanced search

Message boards : Number crunching : To much Work Downloading !!!

AuthorMessage
STE\/E
Send message
Joined: Jul 22 10
Posts: 26
Credit: 38,543,385
RAC: 688
Message 17 - Posted 29 Jul 2010 16:55:58 UTC

    Your Project is messing up my Box's big time, I have my Preferences set to 0.01 Day's to try & Prevent this Project from flooding my Box's with Wu's but even that doesn't work. I went away for a few hour & when I came back I had 2+ Days worth of work on my Box's. So once again I have to Abort all the excess Wu's and wait for the Project to Flood me again & repeat.

    Fabio Campos
    Volunteer moderator
    Project administrator
    Project developer
    Project tester
    Volunteer developer
    Volunteer tester
    Project scientist
    Send message
    Joined: Feb 3 10
    Posts: 65
    Credit: 10,767
    RAC: 11
    Message 18 - Posted 29 Jul 2010 17:45:13 UTC - in response to Message 17.

      Hi PoorBoy,

      I changed some settings: the max_wus_to_send down to 50 and the max_wus_in_progress also to 50 ...

      I hope, it helps ...

      Thx for crunching ;-)

      STE\/E
      Send message
      Joined: Jul 22 10
      Posts: 26
      Credit: 38,543,385
      RAC: 688
      Message 19 - Posted 29 Jul 2010 19:13:38 UTC - in response to Message 18.

        Yes that should help a lot, Thank's

        STE\/E
        Send message
        Joined: Jul 22 10
        Posts: 26
        Credit: 38,543,385
        RAC: 688
        Message 20 - Posted 29 Jul 2010 19:33:26 UTC - in response to Message 19.

          Okay that don't seem to work either, I upped my Contact time to 2 Days on just 1 Box and the Project downloaded 120 Wu's to the Box. Running 4 @ a time and with 1 hour run times only 96 would get done in time with the 1 day deadlines.

          That's not even taking into consideration an other CPU Projects that might what to run during that time. I don't know how others feel but could you lower the settings even more. I'd say 24 in the Cache is fine for me so 24 Max would work for me anyway. That's a 6 hour Chach on my Box's.

          Nasicus
          Send message
          Joined: Jul 18 10
          Posts: 4
          Credit: 2,633,134
          RAC: 10,442
          Message 21 - Posted 29 Jul 2010 19:46:21 UTC - in response to Message 20.

            Last modified: 29 Jul 2010 19:47:49 UTC

            As max WUs in progress I would like to see somthing like about five WUs per core. I think that would be better than xx WUs per host.

            STE\/E
            Send message
            Joined: Jul 22 10
            Posts: 26
            Credit: 38,543,385
            RAC: 688
            Message 22 - Posted 29 Jul 2010 19:49:41 UTC - in response to Message 21.

              Last modified: 29 Jul 2010 19:50:54 UTC

              As max WUs in progress I would like to see somthing like about five WUs per core. I think that would be better than xx WUs per host.


              Yes that would work too ... May even work better as it would take into account the amount of cores you have

              Fabio Campos
              Volunteer moderator
              Project administrator
              Project developer
              Project tester
              Volunteer developer
              Volunteer tester
              Project scientist
              Send message
              Joined: Feb 3 10
              Posts: 65
              Credit: 10,767
              RAC: 11
              Message 23 - Posted 29 Jul 2010 20:23:40 UTC - in response to Message 22.

                Hi,

                so ... I changed again the settings:

                max_wus_in_progress := 20 (per host)
                max_wus_in_progress_cpu := 5 (per core)

                I hope, that's it ...

                Let me know, if it works ...

                Thx again

                STE\/E
                Send message
                Joined: Jul 22 10
                Posts: 26
                Credit: 38,543,385
                RAC: 688
                Message 26 - Posted 29 Jul 2010 20:52:59 UTC - in response to Message 23.

                  Last modified: 29 Jul 2010 20:55:54 UTC

                  Hi,

                  so ... I changed again the settings:

                  max_wus_in_progress := 20 (per host)
                  max_wus_in_progress_cpu := 5 (per core)

                  I hope, that's it ...

                  Let me know, if it works ...

                  Thx again



                  Didn't work, I had 12 Wu's by setting my Contact time to .1 & then I upped the Contact time to 1 Day and the server Sent me a total of 42 Wu's, I upped the Contact time to 2 Days and the servers sent me 34 more Wu's for a total of 76 Wu's.

                  I only have 4 Cores on that Box so all I should be getting is 20 Wu's if your Settings work no matter what I set the Contact time to, at least anything over .2 by my figuring ...

                  Nasicus
                  Send message
                  Joined: Jul 18 10
                  Posts: 4
                  Credit: 2,633,134
                  RAC: 10,442
                  Message 27 - Posted 29 Jul 2010 21:04:01 UTC - in response to Message 23.


                    max_wus_in_progress := 20 (per host)
                    max_wus_in_progress_cpu := 5 (per core)

                    That's ok for a quad core but for computers with more cores it would be better to work only with the second option. At the moment I can use a host with 24 cores so 20 WUs would be too little ;-)

                    Fabio Campos
                    Volunteer moderator
                    Project administrator
                    Project developer
                    Project tester
                    Volunteer developer
                    Volunteer tester
                    Project scientist
                    Send message
                    Joined: Feb 3 10
                    Posts: 65
                    Credit: 10,767
                    RAC: 11
                    Message 28 - Posted 29 Jul 2010 21:11:11 UTC - in response to Message 26.

                      Hmmmmm ...

                      strange ... here is the BOINC definition about those settings:

                      " <max_wus_in_progress> A </max_wus_in_progress>
                      [ <max_wus_in_progress_cpu> B </max_wus_in_progress_cpu> ]
                      [ <max_wus_in_progress_gpu> C </max_wus_in_progress_gpu> ]

                      The maximum jobs in progress on a given host is '''min(A, (B*#CPUs + C*#GPUS)'''.
                      Starting with 6.8, the BOINC client report the resources used by in-progress jobs;
                      in this case, the max CPU jobs in progress is '''B*#CPUs''',
                      the max GPU jobs in progress is '''C*#GPUs''',
                      and the max total jobs in progress is A.
                      "

                      You should get (client version < 6.8) : min(20, (5 * 4 cores)) := 20
                      or (client version >= 6.8) : (5 * 4 cores) := 20


                      I'll setup a client soon to test it ...

                      Thx again ...

                      Fabio Campos
                      Volunteer moderator
                      Project administrator
                      Project developer
                      Project tester
                      Volunteer developer
                      Volunteer tester
                      Project scientist
                      Send message
                      Joined: Feb 3 10
                      Posts: 65
                      Credit: 10,767
                      RAC: 11
                      Message 29 - Posted 29 Jul 2010 21:19:00 UTC - in response to Message 28.

                        @Nasicus

                        Ok,

                        new settings:

                        max_wus_in_progress := 50 (per host)
                        max_wus_in_progress_cpu := 5 (per core)

                        STE\/E
                        Send message
                        Joined: Jul 22 10
                        Posts: 26
                        Credit: 38,543,385
                        RAC: 688
                        Message 30 - Posted 29 Jul 2010 21:20:23 UTC

                          I'm running v6.10.57 on all my Box's, there a Quad Cores except for 1 i7 with HT for 8 Cores ...

                          STE\/E
                          Send message
                          Joined: Jul 22 10
                          Posts: 26
                          Credit: 38,543,385
                          RAC: 688
                          Message 31 - Posted 29 Jul 2010 21:46:38 UTC

                            For Whatever reason that Setting >>> max_wus_in_progress := 50 (per host) doesn't work on my Box's, I can get 200-300 on a Host & Maybe more if I would set the Contact Time High enough but I don't want to have to abort that many off 1 Host.

                            Even with the supposed 50 Limit I get 90+ if I set my Contact Time to 2 Day's, I like a high Contact time for the GPU's because they run the GPU Wu's so fast so you have to keep a High Cache for them but try to restrict the CPU Wu's.

                            But if Nasicus needs more because of his 24 Core Box then by all means give it to him. I'll get my 100,000 here in a few day's and get outta everybody's way ... :)

                            Nasicus
                            Send message
                            Joined: Jul 18 10
                            Posts: 4
                            Credit: 2,633,134
                            RAC: 10,442
                            Message 32 - Posted 29 Jul 2010 22:01:50 UTC - in response to Message 31.

                              But if Nasicus needs more because of his 24 Core Box then by all means give it to him. I'll get my 100,000 here in a few day's and get outta everybody's way ... :)

                              As my box is clearly in minority please take the settings which are ok for most of the crunchers.

                              STE\/E
                              Send message
                              Joined: Jul 22 10
                              Posts: 26
                              Credit: 38,543,385
                              RAC: 688
                              Message 33 - Posted 29 Jul 2010 22:05:53 UTC - in response to Message 32.

                                But if Nasicus needs more because of his 24 Core Box then by all means give it to him. I'll get my 100,000 here in a few day's and get outta everybody's way ... :)

                                As my box is clearly in minority please take the settings which are ok for most of the crunchers.


                                I'm not sure any settings will work on my Box's unless I keep a bare minimum Contact time which I can't do. I'm just stopping New Work on the CPU's & doing a fill up of the GPU Caches every once in awhile and then dropping the Contact time back down again for the CPU's again. That's about all that seems to work for this Project ATM

                                Nasicus
                                Send message
                                Joined: Jul 18 10
                                Posts: 4
                                Credit: 2,633,134
                                RAC: 10,442
                                Message 34 - Posted 29 Jul 2010 22:16:11 UTC - in response to Message 33.

                                  PoorBoy is right, these settings seem not to work as they should. I just got more than 80 WUs on my box which now has over 100 WUs in progress.

                                  Fabio Campos
                                  Volunteer moderator
                                  Project administrator
                                  Project developer
                                  Project tester
                                  Volunteer developer
                                  Volunteer tester
                                  Project scientist
                                  Send message
                                  Joined: Feb 3 10
                                  Posts: 65
                                  Credit: 10,767
                                  RAC: 11
                                  Message 35 - Posted 30 Jul 2010 6:48:53 UTC - in response to Message 34.

                                    Hello again,

                                    I also changed the <daily_result_quota> down to 50 ...

                                    STE\/E
                                    Send message
                                    Joined: Jul 22 10
                                    Posts: 26
                                    Credit: 38,543,385
                                    RAC: 688
                                    Message 36 - Posted 30 Jul 2010 8:52:36 UTC - in response to Message 35.

                                      Hello again,

                                      I also changed the <daily_result_quota> down to 50 ...


                                      Not sure how that will be accepted, my Quads running at their stock 2.4 GHZ can run about 85-90 Wu's Per Day ...

                                      Orakk
                                      Send message
                                      Joined: Jul 30 10
                                      Posts: 2
                                      Credit: 9,173
                                      RAC: 0
                                      Message 37 - Posted 30 Jul 2010 9:14:15 UTC

                                        I have one quad has been given around eighty jobs to do by midnight and it's seven o'clock now.

                                        So what's the story on all these work units will time out in a few hours. Do I reset the project when the clocks goes 31/01/2010?

                                        STE\/E
                                        Send message
                                        Joined: Jul 22 10
                                        Posts: 26
                                        Credit: 38,543,385
                                        RAC: 688
                                        Message 38 - Posted 30 Jul 2010 9:16:50 UTC

                                          I'd just Abort the Excessive Wu's, that's what I've been doing. They'll get sent out right away again or at last they should anyway ...

                                          Orakk
                                          Send message
                                          Joined: Jul 30 10
                                          Posts: 2
                                          Credit: 9,173
                                          RAC: 0
                                          Message 39 - Posted 30 Jul 2010 9:31:21 UTC

                                            Cheers PoorBoy, Shy of any alternative being posted, abort it will be.

                                            Fabio Campos
                                            Volunteer moderator
                                            Project administrator
                                            Project developer
                                            Project tester
                                            Volunteer developer
                                            Volunteer tester
                                            Project scientist
                                            Send message
                                            Joined: Feb 3 10
                                            Posts: 65
                                            Credit: 10,767
                                            RAC: 11
                                            Message 40 - Posted 30 Jul 2010 9:45:09 UTC - in response to Message 38.

                                              Hi,

                                              it seems quite difficult to change the settings, so that everyone agrees with that ...

                                              The settings are now:

                                              <daily_result_quota>
                                              500
                                              </daily_result_quota>
                                              <max_wus_to_send>
                                              50
                                              </max_wus_to_send>
                                              <max_wus_in_progress>
                                              50
                                              </max_wus_in_progress>
                                              <max_wus_in_progress_cpu>
                                              5
                                              </max_wus_in_progress_cpu>

                                              John McLeod VII
                                              Avatar
                                              Send message
                                              Joined: Aug 4 10
                                              Posts: 24
                                              Credit: 521,799
                                              RAC: 377
                                              Message 70 - Posted 4 Aug 2010 2:50:10 UTC - in response to Message 36.

                                                Hello again,

                                                I also changed the <daily_result_quota> down to 50 ...


                                                Not sure how that will be accepted, my Quads running at their stock 2.4 GHZ can run about 85-90 Wu's Per Day ...

                                                Unless things have changed, the daily quota is per CPU and there is an additional amount per GPU (something like 5 X quote / GPU).

                                                John McLeod VII
                                                Avatar
                                                Send message
                                                Joined: Aug 4 10
                                                Posts: 24
                                                Credit: 521,799
                                                RAC: 377
                                                Message 71 - Posted 4 Aug 2010 3:26:20 UTC

                                                  It appears that fpops_est or iops_est is off by a significant margin. This causes an initial download that is vastly too large.

                                                  John McLeod VII
                                                  Avatar
                                                  Send message
                                                  Joined: Aug 4 10
                                                  Posts: 24
                                                  Credit: 521,799
                                                  RAC: 377
                                                  Message 77 - Posted 4 Aug 2010 13:39:04 UTC

                                                    Looks like my duration_correction_factors are coming up between 9 and 23. If everyone else is seeing the same, I would suggest multiplying the fpops_est by something between 10 and 15.
                                                    ____________


                                                    BOINC WIKI

                                                    JAMC
                                                    Send message
                                                    Joined: Oct 18 10
                                                    Posts: 2
                                                    Credit: 28,759
                                                    RAC: 0
                                                    Message 200 - Posted 18 Oct 2010 12:21:59 UTC

                                                      Yikes!!
                                                      I just added this project to run with several other CPU projects on a Q6600 (w/ 2x 4870's, W7_64) and got 50 WU's @ 1 hour/WU, kind of swamping out the other projects... I have cache set at 0.1 day.
                                                      Is this just the way it is here or can I do something to bring the number of downloaded WU's more in line with the other projects?
                                                      Thanks :)

                                                      whynot
                                                      Send message
                                                      Joined: Sep 15 10
                                                      Posts: 30
                                                      Credit: 10,816,644
                                                      RAC: 1,042
                                                      Message 204 - Posted 23 Oct 2010 17:08:08 UTC - in response to Message 200.


                                                        I just added this project to run with several other CPU projects on a Q6600 (w/ 2x 4870's, W7_64) and got 50 WU's @ 1 hour/WU, kind of swamping out the other projects... I have cache set at 0.1 day.


                                                        I've aproximated workload you should have got, and I've summed to 30WU. Thus I overlooked, probably, frequencies of CPU. Then 50WU could be quite OK. And one more note, from my observations, the boinc requests work for each core; thus making you caching for 0.2day instead.


                                                        Is this just the way it is here or can I do something to bring the number of downloaded WU's more in line with the other projects?


                                                        That's the way things work here, indeed. On the matter; no you shouldn't. 'boinc' will learn soon and won't do it again. Moreover, 'primaboinca' is known to checkpoint weirdly. If you have more than 60min for each resume-suspend or if you keep tasks in suspend-to-memory mode you, most probably, wuldn't notice this. If you don't...

                                                        Have fun!

                                                        ____________
                                                        I'm counting for science.
                                                        Points just make me sick.

                                                        Post to thread

                                                        Message boards : Number crunching : To much Work Downloading !!!


                                                        Main page · Your account · Message boards


                                                        Copyright © 2017 primaboinca.com