Message boards : Number crunching : No work - Sprint FB
Message board moderation
Author | Message |
---|---|
Send message Joined: 22 Oct 17 Posts: 3083 Credit: 0 RAC: 0 |
Incredible! The project was left without jobs. Tasks ready to send 0 ice00 maybe you'll run a new experiment? At least in test mode. I have long given you references to assignments 1. wu_500028.rar https://cloud.mail.ru/public/L6yw/2ezKqyCLj 2. wu_519662.ru https://cloud.mail.ru/public/K9kL/DNQqRQavL 3. wu_573392.rar https://cloud.mail.ru/public/GrAJ/uEYmwn5bZ You wrote that the Assimilator and the Validator were prepared. The program for processing jobs is ready. So what's the problem? Why do not you run the experiment? This is a great experiment! I'm doing it manually. I get very interesting solutions! There is no work in the project! I have a lot of work! PS. No, no, the work in the project did not end. Assignments will still be generated. https://yadi.sk/i/S1HhJBwp3Pr9jX |
Send message Joined: 22 Oct 17 Posts: 3083 Credit: 0 RAC: 0 |
ice00 here is the program for processing jobs in my experiment https://progger.info/files/family_mar.exe Author of the program A. Belyshev. Jobs are generated by my program. Of course, tasks are generated by me only for testing. The amount of calculation in the experiment is very large. I hope you run this experiment. https://yadi.sk/i/S1HhJBwp3Pr9jX |
Send message Joined: 7 Jun 18 Posts: 3 Credit: 3,901,681 RAC: 0 |
Incredible! There is a Formula Boinc sprint running at ODLK1 for 3 days: http://formula-boinc.org/sprint.py?sprint=7&year=2018 From the scores so far, the sprint seems to cause about double the participation compared to normal days. But this is only a rough estimation, because the sprint site started counting credits just 6 hours ago. |
Send message Joined: 22 Oct 17 Posts: 3083 Credit: 0 RAC: 0 |
There is a Formula Boinc sprint running at ODLK1 for 3 days: Oh! Great. Thanks for the info. https://yadi.sk/i/S1HhJBwp3Pr9jX |
Send message Joined: 29 Nov 17 Posts: 13 Credit: 2,892,419 RAC: 3,477 |
Double post. Server seems slow, prob due to FB Sprint. |
Send message Joined: 29 Nov 17 Posts: 13 Credit: 2,892,419 RAC: 3,477 |
Enigma was two weeks ago. It has a GPU app but its a major increase in production for a project if there is work. https://stats3.free-dc.org/stats.php?page=proj&proj=eni |
Send message Joined: 28 Oct 17 Posts: 200 Credit: 58,745 RAC: 9 |
Incredible! I did not know about this spring onto ODLK1 before, so server was set optimal with the current past user running it. |
Send message Joined: 29 Nov 17 Posts: 13 Credit: 2,892,419 RAC: 3,477 |
The schedule follows the Formula 1 racing schedule. When there is a race on a weekend there is a Sprint challenge for 3 days. 1 day prior the project is announced. These are the available projects that typically have consistent work. http://formula-boinc.org/sprint.py?lang=&year=2018 |
Send message Joined: 22 Oct 17 Posts: 3083 Credit: 0 RAC: 0 |
It is unfortunate that the project was not ready for competition. Administrators were not able to react quickly and generate new jobs. As a result, the whole competition will pass with no assignments. Many participants of the competition did not receive a single job. I hope that at least after the contest the jobs will be generated and the project will not be stopped. I want to apologize to the participants of the competition. https://yadi.sk/i/S1HhJBwp3Pr9jX |
Send message Joined: 4 Dec 17 Posts: 1 Credit: 5,055,045 RAC: 0 |
No apology necessary. It is very unfortunate that Sebastien, who runs multiple sites, Formula BOINC, WuProp, the LAF forums, etc etc would do this to another project. I really respect what Sebastien does for the BOINC community, but dropping a 4 day major challenge on a project without notifying them is very poor execution of Formula BOINC. Thank you for trying but we understand that without notice there is very little that you can do! |
Send message Joined: 28 Oct 17 Posts: 200 Credit: 58,745 RAC: 9 |
No apology necessary. It is very unfortunate that Sebastien, who runs multiple sites, Formula BOINC, WuProp, the LAF forums, etc etc would do this to another project. I really respect what Sebastien does for the BOINC community, but dropping a 4 day major challenge on a project without notifying them is very poor execution of Formula BOINC. Yes, without a notify I cannot modify actual workunit generation to increase his rate (need to recompile the generator that it is in another work environments actually). Server generates workunits with his buffer that goes empty in less time that was expected. In this case a good solution before starting of competition was to double a workunit task (so one workunit will require double time to get a solution). |
Send message Joined: 7 Jun 18 Posts: 3 Credit: 3,901,681 RAC: 0 |
Yes, it is unfortunate that Formula Boinc (still) runs the sprints without consulting the project admins... Indeed it looks like sprint participants, together with the "baseline" contributors of ODLK1 (mostly Gridcoin), pull the tasks as fast as the project server can generate them. Sprint stats from today, 16:00 UTC = 36 hours into the sprint, show that all teams which are tracked by the Formula Boinc site reached a combined production of 11 million points (that is, 7.3 million per day). This is way up from ODLK1's normal days which see about ~3.3 million points/day. |
Send message Joined: 29 Nov 17 Posts: 13 Credit: 2,892,419 RAC: 3,477 |
Typically on the FB forum there is a thread about the projects available each year. If ODLK, ODLK, or Rakesearch is not want to be included you could ask to be removed. Typically most are included unless there is sporadic work. Even if there is work here, the bandwidth is not keeping up. Best I've seen is 3kb download for the exe files and many attempts on the tasks I do get. A suggestion on this is to make the tasks bigger or bundle them to reduce the constant connections. http://formula-boinc.org/forum/viewtopic.php?f=14&t=100 You could reduce the load by not sending out a 3rd task for a workunit BEFORE the previous deadlines just to later cancel that task. https://boinc.multi-pool.info/latinsquares/workunit.php?wuid=40014965 |
Send message Joined: 28 Oct 17 Posts: 200 Credit: 58,745 RAC: 9 |
Even if there is work here, the bandwidth is not keeping up. Best I've seen is 3kb download for the exe files and many attempts on the tasks I do get. A suggestion on this is to make the tasks bigger or bundle them to reduce the constant connections. Yes, I agree that a task should be make bigger to solve problems. If I did not see wrong, spring project are selected random, so maybe it is for this that no communication is given in time. |
Send message Joined: 22 Oct 17 Posts: 3083 Credit: 0 RAC: 0 |
We have the following results in June for the Application odlk3@home 2018-06-02 01:59 773K 2018-06-03 01:58 815K 2018-06-04 01:59 780K 2018-06-05 01:59 866K 2018-06-06 01:59 905K 2018-06-07 01:59 899K 2018-06-08 01:57 942K 2018-06-09 01:59 1.8M 2018-06-10 01:59 2.1M I processed the results for 2018-06-10, the maximum value is 2.1 M. This gives 24,149 unique CF ODLS! https://yadi.sk/i/S1HhJBwp3Pr9jX |
Send message Joined: 7 Nov 17 Posts: 26 Credit: 13,758,698 RAC: 135 |
I think that you need a simple script. Which tracks the task values. As soon as it fell below a certain value, start generating new jobs. If this happens several times a day, then increase the threshold for generating new jobs. Of course, we need the maximum limits. Then it will work dynamically. |
Send message Joined: 7 Jun 18 Posts: 3 Credit: 3,901,681 RAC: 0 |
The rate at which work was generated was not the only bottleneck, and perhaps not even the most important one. There were periods (and still are) when downloads or uploads or scheduler requests time out, or the scheduler responds with "cannot open database", or the server could not be reached at all. Just stating this as a fact, not as a criticism. Clearly, distributed computing projects cannot hold infinite server and networking resources ready for every perceivable peak load. -------- Today at 04:00UTC, the Formula Boinc sprint finished with these results: http://formula-boinc.org/sprint.py?sprint=7&year=2018 All teams which are tracked by Formula Boinc completed work worth of 21.6 M points together in three days, i.e. 7.2 M / day at average. |
Send message Joined: 22 Oct 17 Posts: 3083 Credit: 0 RAC: 0 |
The rate at which work was generated was not the only bottleneck, and perhaps not even the most important one. There were periods (and still are) when downloads or uploads or scheduler requests time out, or the scheduler responds with "cannot open database", or the server could not be reached at all. Yes, all this is true. The ODLK1 project runs on the personal server ice00. I will note here (not as a reproach, but as a fact) that we were denied help by the server for this project in the Moscow Computer Center. And thank very much ice00 that he provided his personal server for this project. Only thanks to ice00 the ODLK1 project took place. Today at 04:00UTC, the Formula Boinc sprint finished with these results: I think that these are good results. Despite all the difficulties, the competition in the project was successful. https://yadi.sk/i/S1HhJBwp3Pr9jX |
Send message Joined: 22 Oct 17 Posts: 3083 Credit: 0 RAC: 0 |
Along the way ... ODLK and ODLK1 projects are loaded with work completely. I am developing new algorithms. This makes and my colleague Belyshev regardless of me. There are several new experiments. Unfortunately, administrators of ODLK and ODLK1 projects do not want to run new experiments. I am forced to perform experiments manually on my low-power PC. In this regard, I'm looking for a server to run a new branch of the ODLK project - ODLK2. We were much criticized for launching the ODLK1 project. But this project justified our hopes! Now I want to run the ODLK2 project. Criticism is not accepted. The help is needed! Who is ready to help, write, please, PM. https://yadi.sk/i/S1HhJBwp3Pr9jX |
Send message Joined: 28 Oct 17 Posts: 200 Credit: 58,745 RAC: 9 |
The rate at which work was generated was not the only bottleneck, and perhaps not even the most important one. There were periods (and still are) when downloads or uploads or scheduler requests time out, or the scheduler responds with "cannot open database", or the server could not be reached at all. hi, when this happen it is because I'm doing a DB dump every 2 hours for backup porpoise that is transferred then to a removable device. During this, mysql works only for dump to keep integrity safe and so no other requests are done. I still prefer to have often backup that no backup at all :) |
©2024 Progger & Stefano Tognon (ice00) & Reese