Message boards : Technical News : Due to user feedback and an attempt to reduce the load on our servers, we increased the default target run time from 3 to 6 hours, increased job deadlines to 14 days, and added the target cpu runtime option of 2 days
Author | Message |
---|---|
flobee Send message Joined: 28 Jan 19 Posts: 1 Credit: 2,000,189 RAC: 0 |
hmm too bad. means people need to have 24/7 machines. i just tested rosetta the last few days and it gets now only 5% of resources. Reasons: - Too low deadlines (and my cpu time for you is for nothing when the time excceed the deadline? too bad! no, this is not the deal to offer free cpu time and the project dosnt make sense then) - To much disk usage/ Smaller chuncks needed (not possible?) Boinc gets ONLY 3-5GB diskspace. This means it blocks other jobs (like seti) beeing loaded even if 10/12 cores are iddle. they are block by diskspace until one of ~4 rosetta jobs is done to free diskspace. - Note: Only once a day (for 1h) networking is activ. You should calculate how many jobs can be done during the next 24h but: Keep in mind that jobs can iddle for some few weeks because the cpu power is needed by owner. (seti seems to handle this with my setup/setting with enough time buffer to get the jobs done for sure. |
TuxNews Send message Joined: 21 Apr 19 Posts: 2 Credit: 87,761 RAC: 0 |
How to link World Community Grid? i changed account and points are lost... |
Ernext_studio Send message Joined: 28 Mar 20 Posts: 1 Credit: 3,112,191 RAC: 113 |
J'aimerais savoir si c’est possible d'assigné plusieurs core sur la même tâches. La vitesse de travaux fini en temps réel serais maximisé. Pour les tâches non complet ( il devrais avoir une sauvegarde ) au lieu de perdre l'énergie investie sur les death line. |
adrianxw Send message Joined: 18 Sep 05 Posts: 653 Credit: 11,840,739 RAC: 42 |
An observation I have made bothered me. I understand the program generates a random start point, then runs the model to completion. If there is sufficient time left, now knowing the time the model takes on that node, it generates a new random start point and runs again to completion, repeating this until the run time is reached, or at least, running it again would excede the allowed run time. My observation was that if a run of the model fails, ALL the work done in that run is discarded, including, obviously, the earlier, sucessful runs. Wave upon wave of demented avengers march cheerfully out of obscurity into the dream. |
Brian Nixon Send message Joined: 12 Apr 20 Posts: 293 Credit: 8,432,366 RAC: 0 |
Please note that this thread is six years old and the announcement in the OP is no longer applicable. As of mid-2020, default run time is 8 hours, deadlines are 3 days, and the longest run time option is 1½ days. It’s been a long time since secondary-school French, but if I’ve understood your question correctly: I would like to know whether it’s possible to assigned multiple core to the same tasks the answer is no: all Rosetta@home tasks are single-threaded. Rosetta makes use of multi-core CPUs by running multiple independent tasks in parallel. Furthermore, Rosetta tasks run for a fixed duration; they are not seeking to complete a fixed quantity of work in the shortest time possible. Given that a batch of work consists of hundreds of thousands of tasks, minimising the wall time required to complete a single one is not a priority – and a single-threaded implementation is far easier for the programmers to understand and debug. |
mikey Send message Joined: 5 Jan 06 Posts: 1895 Credit: 9,169,305 RAC: 3,857 |
TuxNews asked How to link World Community Grid? i changed account and points are lost... Ask in the Number Crunching thread this one is for Project stuff. |
[AF>France>Est>Alsace]PFLIEGER Guy Send message Joined: 16 Nov 20 Posts: 3 Credit: 602,226 RAC: 0 |
Please note that this thread is six years old and the announcement in the OP is no longer applicable. As of mid-2020, default run time is 8 hours, deadlines are 3 days, and the longest run time option is 1½ days. If you want computing in 2 core then you need to go inside of the boinc manager and put the number of processors to 50% te reason is that you are then running on the full core and not on the thread You are solving in 28000 of second mid computing time with I7 7700 maya2 |
[AF>France>Est>Alsace]PFLIEGER Guy Send message Joined: 16 Nov 20 Posts: 3 Credit: 602,226 RAC: 0 |
this i to show the computing time in full core and not on thread(50% of processors from I7 7700): 1316695629 1178870938 5911174 4 Jan 2021, 18:21:48 UTC 5 Jan 2021, 3:33:40 UTC Terminé et validé 28,799.22 28,748.13 319.92 Rosetta v4.20 windows_x86_64 1316688133 1178864249 5911174 4 Jan 2021, 18:00:38 UTC 5 Jan 2021, 2:22:26 UTC Terminé et validé 28,831.80 28,774.55 458.82 Rosetta v4.20 windows_x86_64 1316684602 1178861066 5911174 4 Jan 2021, 17:51:35 UTC 5 Jan 2021, 1:59:10 UTC Terminé et validé 28,939.51 28,891.30 380.77 Rosetta v4.20 windows_x86_64 1316675672 1178852861 5911174 4 Jan 2021, 17:29:57 UTC 5 Jan 2021, 1:40:20 UTC Terminé et validé 28,570.94 28,521.39 467.47 Rosetta v4.20 windows_x86_64 1316510423 1178704370 5911174 4 Jan 2021, 10:19:53 UTC 4 Jan 2021, 19:33:40 UTC Terminé et validé 28,843.99 28,801.56 322.71 Rosetta v4.20 windows_x86_64 1316535327 1178696564 5911174 4 Jan 2021, 10:03:59 UTC 4 Jan 2021, 18:21:48 UTC Terminé et validé 28,784.77 28,735.95 351.46 Rosetta v4.20 windows_x86_64 1316498017 1178693069 5911174 4 Jan 2021, 9:45:49 UTC 4 Jan 2021, 18:00:38 UTC Terminé et validé 28,366.14 28,312.59 526.36 Rosetta v4.20 windows_x86_64 1316490887 1178686585 5911174 4 Jan 2021, 9:25:29 UTC 4 Jan 2021, 17:51:35 UTC Terminé et validé 28,819.22 28,768.98 465.32 Rosetta v4.20 windows_x86_64 1316336198 1178546337 5911174 4 Jan 2021, 2:26:23 UTC 4 Jan 2021, 11:32:28 UTC Terminé et validé 28,843.32 28,781.77 454.75 Rosetta v4.20 windows_x86_64 maya2 |
Falconet Send message Joined: 9 Mar 09 Posts: 353 Credit: 1,227,479 RAC: 1,235 |
The computing time won't change since it's a fixed runtime based on whatever your preference is on the preferences page. What will probably happen is that more models are completed within the same runtime (8 hours in your case) since you are using just the cores that are somewhat faster than running all 8 threads. |
Brian Nixon Send message Joined: 12 Apr 20 Posts: 293 Credit: 8,432,366 RAC: 0 |
you are then running on the full core and not on the threadNone of that user’s machines has Hyper-Threading. |
Falconet Send message Joined: 9 Mar 09 Posts: 353 Credit: 1,227,479 RAC: 1,235 |
you are then running on the full core and not on the threadNone of that user’s machines has Hyper-Threading. Hmm, he mentioned an I7-7700 which is a 4C/8T CPU which is on the list: https://boinc.bakerlab.org/rosetta/hosts_user.php?userid=2296186 Your link is not from the same user but rather from another user up on this page :) |
Brian Nixon Send message Joined: 12 Apr 20 Posts: 293 Credit: 8,432,366 RAC: 0 |
Indeed. But maya2 was replying to me replying to lw elite, for whom advice to reduce to 50% of CPUs would be inappropriate. Maybe something got lost in translation. They could just cut out the middle man, and speak to each other in French… :-) (FWIW I tried restricting to 50% of my Hyper-Threaded CPUs a while ago; I didn’t record the results but IIRC I got about 80% more credits per task.) |
Luc Lefebvre Send message Joined: 9 Dec 05 Posts: 5 Credit: 1,160,134 RAC: 0 |
I will stop to run Rosetta I am running it since years but unfortunately I have really big issues to deliver any working unit now due to the little time aloted for us to do the computing. I am running multiple projects and can't abrely deliver any roseta in time. It is annoying to get a unit at 70% and to see that you can't deliver it in the few hours left still available to deliver it. If you can give us a few weeks to deliver a unit like ofther projects I might come back. Sorry about that but tired of loosing computer power for a unit that wont be use at the end. |
Luc Lefebvre Send message Joined: 9 Dec 05 Posts: 5 Credit: 1,160,134 RAC: 0 |
Certificate of Computation This certifies that Luc Lefebvre has participated in Rosetta@home since 9 December 2005,and has contributed 1,160,134 Cobblestones of computation (1.00 quintillion floating-point operations)to Rosetta@home. |
Jim1348 Send message Joined: 19 Jan 06 Posts: 881 Credit: 52,257,545 RAC: 0 |
I will stop to run Rosetta I am running it since years but unfortunately I have really big issues to deliver any working unit now due to the little time aloted for us to do the computing. I keep the default buffer of 0.1+0.5 days (sometimes 0.25+0.5 days when running other long projects) and don't see the problem. I get the occasional high priority, but they complete on time. I have added several big machines recently (Ryzen 3900X, 3950X), and might do a bit more. EDIT: Yes, they run 24/7. You need that for Rosetta now. |
Luc Lefebvre Send message Joined: 9 Dec 05 Posts: 5 Credit: 1,160,134 RAC: 0 |
MY pc's only running a few hours a day not 7/24 and due to the nature of my work I can't allot more then 25% of my cpu power. Will put focus on others projects that I can deliver on a timely manner then. Tx for the info well appreciated. |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1681 Credit: 17,854,150 RAC: 22,647 |
MY pc's only running a few hours a day not 7/24 and due to the nature of my work I can't allot more then 25% of my cpu power. Will put focus on others projects that I can deliver on a timely manner then. Tx for the info well appreciated.The other option is to change the Target CPU time to a value less than 8 hours and run with no cache. Grant Darwin NT |
Johannes Kingma Send message Joined: 4 Dec 17 Posts: 33 Credit: 9,453,639 RAC: 0 |
As I use my computer as heating It is fine for me to run it 24/7. Energy is cheap hear and I can lower the central heating. So go ahead. |
mikey Send message Joined: 5 Jan 06 Posts: 1895 Credit: 9,169,305 RAC: 3,857 |
As I use my computer as heating It is fine for me to run it 24/7. Energy is cheap hear and I can lower the central heating. So go ahead. And here I pay to keep my pc room cool!!! I run my pc's 24/7 as well |
billy ewell 1931 Send message Joined: 30 Mar 07 Posts: 14 Credit: 6,899,522 RAC: 0 |
It is absolutely beyond my computing intelligence to remotely understand what is going on with the processing of PYTHON tasks.It is inconceivable that 50-51 seconds of computing time is needed to achieve 1 second reduction in time remaining. At that rate of processing virtually over 90% or so of downloaded tasks will be declared "no response" because they exceeded the time limit for reporting. Apparently I will disconnect from Rosetta after I await an overnight response from someone that can enlighten me as to what is the problem Surely I am doing something wrong. |
Message boards :
Technical News :
Due to user feedback and an attempt to reduce the load on our servers, we increased the default target run time from 3 to 6 hours, increased job deadlines to 14 days, and added the target cpu runtime option of 2 days
©2024 University of Washington
https://www.bakerlab.org