Message boards :
Number crunching :
cancel wu's
Message board moderation
Author | Message |
---|---|
Send message Joined: 21 Aug 19 Posts: 5 Credit: 4,991,708 RAC: 33,966 |
hello is it normal to cancel wu's already partially calculated without crediting them? thank you 18 Oct 2020, 15:32:00 UTC 18 Oct 2020, 15:58:35 UTC Annulé par le serveur 1,183.18 1,171.00 --- 2_Gaia@home v1.00 18 Oct 2020, 15:00:51 UTC 18 Oct 2020, 16:46:25 UTC Annulé par le serveur 6,316.74 6,126.78 --- 2_Gaia@home v1.00 18 Oct 2020, 15:32:00 UTC 18 Oct 2020, 15:58:35 UTC Annulé par le serveur 1,183.18 1,171.00 --- 2_Gaia@home v1.00 |
Send message Joined: 21 Aug 19 Posts: 23 Credit: 91,637 RAC: 0 |
In the next topic, the admin gave the answer Staram sie w chwili zaistnienia problemu wywolywac przerwanie obliczen przez serwer w celu ochrony czasu obliczen na Panstwa procesorach. At the moment of the problem, I try to cause the server to interrupt the computation in order to protect the computing time on your processors. |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
The queue is cleared for tasks where the given number of calculations is achieved for some targets ( about 1 % tasks). I will create automatically system for add jobs for calculations respect achieved solutions, without remove jobs form queue. We plan to send about 750 000 wus for calculations ... |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
queue cleaning complete |
Send message Joined: 21 Aug 19 Posts: 23 Credit: 91,637 RAC: 0 |
We plan to send about 750 000 wus for calculations ... Quite a large amount of calculations, accordingly, statistics of received credits ;o) I would like to hear a small comment about the structure of the planned calculations Will this be one app (current version)? Or do new applications / versions appear as calculations are made? Will the statistics be common for the entire project or are you planning to account separately for applications? If the "reward" for the amount of computation has already been determined, are you planning on introducing a badge system to reward users who reach certain milestones? It is advisable for participants to plan ahead of time for the opportunity to achieve certain goals, both locally in the project and on external statistics sites |
Send message Joined: 29 Sep 20 Posts: 14 Credit: 64,341 RAC: 0 |
At the moment of the problem, I try to cause the server to interrupt the computation in order to protect the computing time on your processors.This is not happening with this task on my machine: Application 2_Gaia@home 1.00 Name 2_7281 State Running Received Sat 17 Oct 2020 13:38:59 CEST Report deadline Mon 19 Oct 2020 13:38:58 CEST Estimated computation size 3,600 GFLOPs CPU time 1d 01:46:33 CPU time since checkpoint 1d 01:46:33 Elapsed time 1d 01:46:35 Estimated time remaining 00:00:00 Fraction done 100.000% Virtual memory size 11.54 MB Working set size 8.90 MB Directory slots/1 Process ID 4876 Progress rate 3.960% per hour Executable 2_Gaia@home[20201017.07]_x86_64-pc-linux-gnu and it's not (yet) aborted by the server. |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
2_Gaia@home : The 2_Gaia@home check system time before start main loop of calculation. Then, it checks the system time on each loop steps. The main loop is broken if time difference between start and actual time is greater than 2h. The 2_Gaia@home is finishing work and prepare results. The number of lines in the output file is different for different processors. Also, a surprise for me is why in some cases the loop does not end. :( These are sporadic cases. |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
I think I found out why this is happening with 2_Gaia@home We numerically calculate the motion of the star cluster near the Sun in the gravitational field of the Galaxy. For each star of cluster we draw clones using the covariance matrix from the Gaia catalog. Sometimes a random clone requires a very small integration step which increases the computation time for the loop step (2_Gaia@home app). Unfortunately, we are not able to predict such a situation :( |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
Status of apps: 1_Gaia@home - final 2_Gaia@home - final (I hope, We will check the received numbers for the first full results obtained today. Random checks were positive) Hopefully the new 2_Gaia@home computing strategy will benefit you and us. (If not then I will look for the next calculation strategy) > I would like to hear a small comment about the structure of the planned calculations Currently, all calculations are performed using the Gaia DR2 star catalog. 1_Gaia@home calculates the trajectories of comet clones in the gravitational field of the star cluster and the galaxy 2_Gaia@home calculates the movement of star clones in the gravitational field of the galaxy. The DR3 Gaia catalog will be available in a few months, so we will start new calculation with new data (using 1_Gaia@home and 2_Gaia@home). We would also like to start calculating very high precision quadrupole calculations (3_Gaia@home) Future, the aim of the project is to enable the use of boinc for scientists who calculate using the Gaia catalog. Since each topic has its own specificity, it requires a lot of work to be implemented efficiently. I hope that the stable versions of the existing calculations will compensate for testing new issues... >Will the statistics be common for the entire project or are you planning to account separately for applications? >If the "reward" for the amount of computation has already been determined, are you planning on introducing a badge system to reward users who reach certain milestones? >It is advisable for participants to plan ahead of time for the opportunity to achieve certain goals, both locally in the project and on external statistics sites Unfortunately, I didn't have time to go deeper into the documentation on this topic. You know what solution is best for you and I am asking you for support. I don't want to create this badge system without knowing it. |
Send message Joined: 21 Aug 19 Posts: 23 Credit: 91,637 RAC: 0 |
It would be nice to move the last 2 posts to a separate topic in the "Science" And also add information about other applications to the "About Us" page. Or replace with something more general + description in Science About badges started a separate topic |
Send message Joined: 29 Sep 20 Posts: 14 Credit: 64,341 RAC: 0 |
2_Gaia@home : I think I found out why this is happening with 2_Gaia@home So, for us cruchers it's OK to abort tasks running longer than ~3 hours, cause when running longer, you don't get a valid result and we will not get credit for the wasted time. Problem is that such a task would be sent to another 'victim', so maybe a temporary solution to reduce wasted time (until you have a better solution within your application) is to reduce the rsc_fpops_bound from 86400000000000 to 21600000000000. |
Send message Joined: 21 Aug 19 Posts: 23 Credit: 91,637 RAC: 0 |
reduce the rsc_fpops_bound from 86400000000000 to 21600000000000 this could potentially reduce the size of credits from ~64 to ~16 you need to check how Credit-New will behave |
Send message Joined: 15 Oct 19 Posts: 11 Credit: 2,848,916 RAC: 0 |
No project should be using credit-new....it is so prone to error and wonky results. |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
I will try to change 2_Gaia@home like this: I will use the kernel signal to terminate after 3h and I will try to save some temporary results so that the program exits properly and you won't lose your credits. I hope I can do it ... what do you think about it ? |
Send message Joined: 29 Sep 20 Posts: 14 Credit: 64,341 RAC: 0 |
I will try to change 2_Gaia@home like this: You could give that a try and I hope those temporary results are still useful for you. |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
I start cleaning the queue and wait for wus in progress |
Send message Joined: 12 Oct 20 Posts: 12 Credit: 1,558,306 RAC: 2,243 |
sounds good to me, I'm back to do work |
Send message Joined: 21 Aug 19 Posts: 110 Credit: 888,695 RAC: 8,066 |
3_Gaia@home - test for new vesrion of 2_Gaia@home (350 wus) (normal time of calculation: 1h, stop signal 1,5h) |
Send message Joined: 29 Sep 20 Posts: 14 Credit: 64,341 RAC: 0 |
3_Gaia@home - test for new vesrion of 2_Gaia@home (350 wus) (normal time of calculation: 1h, stop signal 1,5h)Let's see how these 360 workunits behave. I've 16 tasks running and 6 ready to start. |
Send message Joined: 29 Sep 20 Posts: 14 Credit: 64,341 RAC: 0 |
Results are not reporting their used CPU-time. |
©2024 GAVIP-GC