Message boards : Number crunching : Problems and Technical Issues with Rosetta@home
Previous · 1 . . . 56 · 57 · 58 · 59 · 60 · 61 · 62 . . . 293 · Next
Author | Message |
---|---|
Mr P Hucker![]() Send message Joined: 12 Aug 06 Posts: 1600 Credit: 11,231,553 RAC: 1,264 ![]() |
"Deadlines for Rosetta@home tasks are 3 days from the time they are issued. " Perhaps the scientists don't want to wait two weeks to get the answers they need? That seems perfectly reasonable to me. A task takes 8 hours, so 3 days is plenty of time. Your Boinc client shouldn't be downloading more than it can get done. It should learn how often your computer runs, what other projects you're doing, etc. But the best way is to set your cache to a small value - I have mine on 0.13 + 0.13 days (which is 3-6 hours). There's no need to queue lots of tasks. ![]() |
KWSN Ekky Ekky Ekky Send message Joined: 3 Apr 20 Posts: 9 Credit: 5,062,511 RAC: 0 |
The Rosetta deadlines are beyond silly. People are crunching and losing time after time. I am not complaining about work that take "two weeks" but three days. I keep one alleged day of work but when Rosetta's unrealistic times are always vastly shorter than the actual time it takes, it is ridiculous. Then to reject a result because it is out of such a daft short time is to be frankly insulting. I do note the very large number of people who have left and continue to leave the project and it does not surprise me in the least. It has to be time to rething some things and that includes the three day time limit. |
Mr P Hucker![]() Send message Joined: 12 Aug 06 Posts: 1600 Credit: 11,231,553 RAC: 1,264 ![]() |
The Rosetta deadlines are beyond silly. I don't understand what you're getting at. You receive a task that takes 8 hours to complete, and you have to send it back in 3 days (9 times longer than it takes to do it). How can that possibly cause you not to get them done in time? The predicted times are not "unrealistic", they're precisely correct. In fact they're somehow limited to 8 or 12 hours. If yours take longer than that, there's something very wrong with your computer. And it's not a daft short time, it's presumably because the scientists need those results back quickly so they can work on the next stage. I've looked at your computers, and two of them are working just fine, all tasks were validated fine, and done in exactly 8 hours each. Only the i5-8400 is playing up, it's taking 24 hours to do a task, but only doing 8 hours of processing. Is your CPU doing something else aswell? Is it overheating? Have you limited Boinc to use 33% of CPU time? If the latter, change that to 100% and put the cores down to 33% instead. That will run less tasks at once, but at full speed. ![]() |
Daedalus Send message Joined: 1 Aug 08 Posts: 39 Credit: 10,066,506 RAC: 63 |
The BOINC manager is not that good at predicting my compute time and anyway i am not home on weekends so i always have some tasks due to finish a day my computer will be off. So at the end of the week i have to cancel a batch. Systematically. I understand they want their results fast but i cannot always provide. ![]() |
![]() Send message Joined: 16 Jun 08 Posts: 1227 Credit: 14,140,275 RAC: 238 |
The BOINC manager is not that good at predicting my compute time and anyway i am not home on weekends so i always have some tasks due to finish a day my computer will be off. So at the end of the week i have to cancel a batch. Systematically. Have you tried setting No new tasks about the time any more tasks downloaded will not finish in time? |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2059 Credit: 40,463,815 RAC: 1,453 ![]() |
"Deadlines for Rosetta@home tasks are 3 days from the time they are issued." Not sure if you count it as a sensible reason, but around 6,000 people are dying in the world of CV19 every day for the last few months. The quicker they come back, the quicker the next iteration of tasks can go out again. You're not spending a lot of time crunching tasks if you can't get an 8hr task back within 3 days. The average turnaround time at this project is 1.3 days, including all delays - a very large number inside 24hrs. Either you're machine isn't working on them much or your buffer of tasks is too high to work through them by the deadline - or a combination of the two. Especially while tasks are in short supply, you should reduce your offline cache of tasks by a margin that allows you to complete them on time instead of just waiting in a queue, because many others have run out of work and would complete them inside half a day. You picked the wrong day to say you've got too many tasks to complete within the deadline... ![]() ![]() |
Daedalus Send message Joined: 1 Aug 08 Posts: 39 Credit: 10,066,506 RAC: 63 |
Yes but it doesn't work out well enough. Reducing the cache helped better but i still have to cancel some. It doesn't help that i sometimes stop to crunch for Folding. :) |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2059 Credit: 40,463,815 RAC: 1,453 ![]() |
I do note the very large number of people who have left and continue to leave the project and it does not surprise me in the least. Before Seti shut down there were 40-50k hosts returning tasks daily here. Currently it's around 550k daily. Good observation. After the massive influx of new users and hosts, the project did have a massive rethink, doubling the minimum runtime and cutting the deadline from 8 days to 3. It was massively successful, enabling 10x as many tasks to be returned with more useful information in a lot less time, while not exceeding the server capacity (which was being hammered at the time) That being the case, if you could tweak your settings appropriately to your ability to run and meet the deadline the project requires, that would be great. Or you could continue being late to return your tasks and get no credit for them. Your choice. ![]() ![]() |
![]() Send message Joined: 16 Jun 08 Posts: 1227 Credit: 14,140,275 RAC: 238 |
Folding@home? If so, I have Folding@home doing GPU work only, with BOINC projects taking up all of the CPU time they can without interfering with Folding@home or with my email reading. That allows both Folding@home and BOINC to crunch nearly 24 hours a day. |
Sid Celery Send message Joined: 11 Feb 08 Posts: 2059 Credit: 40,463,815 RAC: 1,453 ![]() |
Have you tried setting No new tasks about the time any more tasks downloaded will not finish in time? I have no idea as I've never used it, but would it help to set something in Boinc's "Daily Schedules" tab under Computing Preferences? I wonder what it would do. ![]() ![]() |
Daedalus Send message Joined: 1 Aug 08 Posts: 39 Credit: 10,066,506 RAC: 63 |
I don't know. It doesn't seem to forbid specific days. Just specific hours some days. In the meantime, i will try to reduce further the cache to see if it helps well enough. :) |
![]() Send message Joined: 16 Jun 08 Posts: 1227 Credit: 14,140,275 RAC: 238 |
That appears to mean that you can exclude certain days by excluding all of the hours on those days. I'd expect how much it helps to depend on how well BOINC uses the excluded days in its calculations of how much work to download. |
Mr P Hucker![]() Send message Joined: 12 Aug 06 Posts: 1600 Credit: 11,231,553 RAC: 1,264 ![]() |
The BOINC manager is not that good at predicting my compute time and anyway i am not home on weekends so i always have some tasks due to finish a day my computer will be off. So at the end of the week i have to cancel a batch. Systematically. Just leave the machine on. Computers enjoy crunching :-) I've got 6 running 24/7. 5 of them in another room where I can't hear the fans! ![]() |
Arnav Sood Send message Joined: 20 Aug 18 Posts: 2 Credit: 11,782,086 RAC: 0 |
I've been seeing a lot of error while downloadingtasks, and the website sometimes gives me the message "Project is Down." Is this just me or are others seeing this too? |
Jim1348 Send message Joined: 19 Jan 06 Posts: 881 Credit: 52,257,545 RAC: 0 |
Is this just me or are others seeing this too? Me too. https://boinc.bakerlab.org/rosetta/results.php?hostid=4684790 |
2fifty6 Send message Joined: 24 Mar 20 Posts: 2 Credit: 946,426 RAC: 0 |
In this era of high speed, always-on internet connections, what purpose does the cache even serve? I have mine set to zero so it just downloads WUs as others finish and it works just fine. And since those WUs are always "fresh," I never have to worry about running up against the deadline. |
Mr P Hucker![]() Send message Joined: 12 Aug 06 Posts: 1600 Credit: 11,231,553 RAC: 1,264 ![]() |
I run Milkyway, where the tasks take one minute on a graphics card. I can't download them that often, the server doesn't allow it. I only set my cache to 3+3 hours though. For example, on my CPUs I currently have 66 CPU tasks running and 12 tasks queued. It doesn't seem right to me to be constantly downloading 1 task at a time, I'm sure the server prefers I download batches of say 6. In fact on my 24 core machines, they could be trying to get tasks from Rosetta extremely often if they got one at a time. What if one task finished at 3:00pm, so it downloaded a new task, then another core finished at 3:01pm? And another at 3:02pm? I suspect I'd hit a server limit of how often I can contact. ![]() |
2fifty6 Send message Joined: 24 Mar 20 Posts: 2 Credit: 946,426 RAC: 0 |
I run Milkyway, where the tasks take one minute on a graphics card. I can't download them that often, the server doesn't allow it. I only set my cache to 3+3 hours though. For example, on my CPUs I currently have 66 CPU tasks running and 12 tasks queued. It doesn't seem right to me to be constantly downloading 1 task at a time, I'm sure the server prefers I download batches of say 6. In fact on my 24 core machines, they could be trying to get tasks from Rosetta extremely often if they got one at a time. What if one task finished at 3:00pm, so it downloaded a new task, then another core finished at 3:01pm? And another at 3:02pm? I suspect I'd hit a server limit of how often I can contact.Ok, that sounds like a reasonable fringe case. But for the vast majority of people whose computers take hours to run each WU, it doesn't seem like a cache of jobs really adds much benefit from a project standpoint. It just unnecessarily delays WU completion as they sit claimed-but-idle in someone's queue. |
Mr P Hucker![]() Send message Joined: 12 Aug 06 Posts: 1600 Credit: 11,231,553 RAC: 1,264 ![]() |
Ok, that sounds like a reasonable fringe case. But for the vast majority of people whose computers take hours to run each WU, it doesn't seem like a cache of jobs really adds much benefit from a project standpoint. It just unnecessarily delays WU completion as they sit claimed-but-idle in someone's queue. Agreed, for normal computers without Milkyway on GPU, or with huge numbers of cores. A long time ago I'd use a big cache on projects which had a lot of server downtime, but I don't think many do that much now. Even if they did, if you have two projects your computer can fall back on the other one. But a queue of 3+3 hours (I think the Boinc default is pretty similar) isn't much, since it also includes the time for the running tasks to complete. Say you have 4 cores, and they have all recently started an 8 hour Rosetta task, they all have 7 hours to completion. No tasks will be queued, because Boinc can see all 4 cores are going to be busy for 7 hours. Only when it gets below 3 hours will it download anything. At this point with my setting, it would get one task per core, and if you have 24 cores like two of my machines do, zero queue would mean constantly hammering the server for work. ![]() |
IT_POWER_SALZKOTTEN Send message Joined: 19 Feb 20 Posts: 1 Credit: 1,011,511 RAC: 279 ![]() |
i don't get new work, too, Some days ago the "NumberFields@home" had a problem: Expired SSL certificates. User action was required. Maybe here the same? |
Message boards :
Number crunching :
Problems and Technical Issues with Rosetta@home
©2024 University of Washington
https://www.bakerlab.org