I am so sorry to quit this project, but it is always too late to report the result because the estimated time is always larger than 5 hours, while my 6 core computer is always free of other jobs and calculating with 11+ GB of Rosetta data on the disk.

I'm not really satisfied with the huge packets because I don't arrive at time schedule for reporting.
Therefore I quid the project.

World Community grid always sends little packets with average 2h caluclating time, and that's always done before deadline...

If you set your cache to 1 day (o better yet less) then you wouldn't miss any deadlines.

---

Grant
Darwin NT

I'm new here, too, Wim but I have always had my cache set to 0.3 days and 0.3 additional so that whichever project I have contributed to (currently only here and Folding), I only get 1 task per allowed core plus maybe a couple spare so I never have more work than I can do by deadline.
And there's also the option in Preferences to set shorter WUs down as low as 2hrs.
Hope that'll change your mind on quitting.

Yes, in BOINC Manager, select View: Advanced Mode.
Then go to Options, Computing Preferences...
Use small numbers for the first two "Other" options (Store at least n1 days work, store up to n2 additional days work), and you won't get a bunch of tasks getting old while waiting to run. As Ray suggested, try n1=0.3, n2=0.3.
And, of course, make sure your computer isn't going to sleep while processing. Turning off the display is fine, but don't suspend the whole system.

Yes, in BOINC Manager, select View: Advanced Mode.
Then go to Options, Computing Preferences...
Use small numbers for the first two "Other" options (Store at least n1 days work, store up to n2 additional days work), and you won't get a bunch of tasks getting old while waiting to run. As Ray suggested, try n1=0.3, n2=0.3.
And, of course, make sure your computer isn't going to sleep while processing. Turning off the display is fine, but don't suspend the whole system.

Or just use the web based preferences on your account page.

---

Grant
Darwin NT

I too am sorry having to leave Rosetta@Home. Two months ago I signed up to R@H to support the search for a Covid solution. I installed it on 3 of my computers and everything had been running fine until a change was implemented a couple of weeks ago. The change was that the deadline is now just 3 days instead of the previous 7 days. Two of the three computers are (somewhat) weak (but run for many hours a day), so I had to adjust the settings so that work is only done 50% percent of the time on both computers. On one computer (Intel Core i3-4030U) I also had to change the settings as to not run work while the computer is in use and on the other of the two (Intel J3160) I had to limit work to 2 of the 4 cores. That was no issue when the deadline was at 7 days. The third computer is more powerful (Intel G4400) but only runs for 2-3 hours per day. Again, with the 7 day deadline this was no issue.

Immediately after the deadline had been changed to 3 days I already had to stop work on 2 computers because they could never reach the deadline and today I also had to disconnect the 3rd computer. Recently work has been sent (Junior_HalfRoid_design5_COVID-19 tasks) that take much longer than the previous tasks. They too do not end on time (the other tasks did).

My local settings are 0.1/0.5 days, the default settings in BOINC. I therefore never have excess WU and they usually started quickly after they arrived.

So it is with great sadness that I cannot participate anymore, unless sometime in the future the deadline will be longer again. I was really happy to be able to participate, but it is impossible at the moment. That is why I now take the effort to write this down, because it really is a disappointment for me having to stop contributing.

I do not know whether this is the correct forum to post this message or that Number crunching is a better place. My main goal for posting is to inform the developers that a 3 day deadline is too short. I think for R@H it is a pity to lose people who are willing to contribute but can’t do that anymore.

My local settings are 0.1/0.5 days,

You would be better off with 0.5 & 0.02 days if you wanted to carry half a days worth of work. Running more than one project, 0.05 & 0.01 would be better still.

My main goal for posting is to inform the developers that a 3 day deadline is too short.

The default target CPU Runtime is 8 hours. To do 8 hours work over 3 days is not a big ask or expectation.

If a system is only on for a few hours per day, then selecting a 2 hour Target CPU time with no cache would still allow you to do work, and not miss deadlines (some Tasks may take longer than 2 hours before they are able to finish & produce a Valid result).
Also, changing some of the defaults of when BOINC can process work would allow the systems to do considerably more while they are on. The defaults are excessively conservative- Rosetta runs at Idle priority & pretty much everything else tends to have a higher priority so pausing BOINC when the system is in use or non-BOINC CPU usage is occurring just isn't necessary in 99% of cases.

---

Grant
Darwin NT

Your average turnaround time is 2 days. And they've just taken steps to help ensure machines are not sent more work than they can complete before the deadline. They've also just implemented changes to do more checkpointing on long-running models, so if the daily shutdown was losing progress on the work unit, that issue has now been addressed as well.

I am always frustrated when people take their first post on the message boards to tell me they are in some way fed up and leaving. You've got a thousand people here that have run in to just about any circumstance you can imagine, and you chose to use the communication of the message boards only to say good bye, rather than ask questions. Please ask questions.

---

Rosetta Moderator: Mod.Sense

Your average turnaround time is 2 days. And they've just taken steps to help ensure machines are not sent more work than they can complete before the deadline. They've also just implemented changes to do more checkpointing on long-running models, so if the daily shutdown was losing progress on the work unit, that issue has now been addressed as well.

I am always frustrated when people take their first post on the message boards to tell me they are in some way fed up and leaving. You've got a thousand people here that have run in to just about any circumstance you can imagine, and you chose to use the communication of the message boards only to say good bye, rather than ask questions. Please ask questions.

Thank you for your response and my apologies for your frustration. I was not saying I was fed up, but I was unable to run R@H anymore under the current circumstances and frustrated because of that fact. That is why I took the effort to post here instead of just leaving. Please consider that a sign of my involvement. Also note that I have been running R@H for about a month on my 3 computers without any issue. But now that the deadline was shortened, on 2 of my computers it was immediately not possible anymore to return WUs on time and on my 3rd computer the problems started with the new Junior_HalfRoid_design5_COVID-19 tasks that apparently take twice the time the other tasks did. I also said that if in the future the deadline would be longer again, I would participate again. So I am giving a solution and not saying I am fed up.

Maybe there is a good reason why the deadline has changed from 7 to 3 days, but I have not read it and I did try to find it. To me a deadline of 7 days instead of 3 days seems the simplest solution. I noticed it that these tasks were not completed on time anymore, because I occasionally check the clients. If I had not ,looked BOINC would have been running without being able to return work, without me knowing it. That is why I was surprised the deadline had been shortened and especially that there was no message about that in the BOINC client.

What questions would you like me to ask? I have tried to find answers on the forum by reading various posts, but could not find any solution. For me it was the shortening of the deadlines that was/is the problem. And TBH IMHO I should not have to tweak BOINC too much, especially as a newbie to the project. I also run World Community Grid and I did not have to tweak anything for that. Tasks take less time and the deadline is longer, therefore their deadlines are always met.

Are you saying that with more checkpointing I should not receive the Junior_HalfRoid_design5_COVID-19 tasks anymore? I do not know exactly how to interpret your remark. I think you are saying that more checkpointing should prevent losses from shutting down, because there is less time between those checkpoints and therefore losses are smaller. I am going to try whether that solves not being able to finish the work on time, even though I have a bit of doubt whether that will solve it, because the computer is shut down only once per day and at the deadline the task always was only around 50-60% completed. I read somewhere else that BOINC does not stop working on a WU after the deadline has passed, because some projects will extend it if the computer still works on it, and other projects only have a deadline as a formality. However, in all cases where I did not meet the deadline for the Junior_HalfRoid_design5_COVID-19 tasks I could see that the task was distributed again within 5 minutes to other far more powerful computers and I was 100% certain that they would beat me if I kept the WU going. Is there any way to not receive the longer WU's? I am still rather new to the topic of project and BOINC settings, so I do not know.

You also say that machines are not sent more work than they can complete before the deadline. Does this mean the quantity of the work or also the size? Usually all work was started very quickly after receiving it, so on my side it was not a problem of having too many items in "stock".

Despite what you write, I still cannot run R@H anymore on my other 2 computers, because they were not able to meet the new 3 day deadline for smaller tasks. One of those computers only runs for 2-3 hours per day, the other has a weak processor (J3160) for which I had to include a couple of processing restrictions (see my previous post). It seems here the only solution can only be a longer deadline, hence my point for the developers to increase the deadline.

You write that my average turnaround time is 2 days. Where can I find that? I have not found it in the client or in my account, but I am likely overlooking it. I suspect these 2 days only relate to the one computer on which I had still been running R@H.

Thanks in advance for your reply.

Pim

Despite what you write, I still cannot run R@H anymore on my other 2 computers, because they were not able to meet the new 3 day deadline for smaller tasks. One of those computers only runs for 2-3 hours per day, the other has a weak processor (J3160) for which I had to include a couple of processing restrictions (see my previous post). It seems here the only solution can only be a longer deadline, hence my point for the developers to increase the deadline.

No, the other solution that i pointed out & you seem to have missed is that you can set the Target CPU Runtime to a value that is less than the 8 hour default, and set a very small cache size (with more than one project, no cache is best). And you can change the settings on your system so Rosetta is able to do more work while the system is running without impacting on it's general performance/responsiveness.

IMHO I should not have to tweak BOINC too much, especially as a newbie to the project.

BOINC allows multiple projects to be run on a single (or multiple) systems. The default settings will work for pretty much any project & system, they are set for the vast majority of users & projects. However for edge cases such as yours, and when more than one project is being done, changes to the defaults will be required to meet those particular circumstances.

You write that my average turnaround time is 2 days. Where can I find that?

In your Account, Computers, Details. Also per application- Application details, Show.

---

Grant
Darwin NT

No, the other solution that i pointed out & you seem to have missed is that you can set the Target CPU Runtime to a value that is less than the 8 hour default, and set a very small cache size (with more than one project, no cache is best). And you can change the settings on your system so Rosetta is able to do more work while the system is running without impacting on it's general performance/responsiveness.

Thanks for your replies. I did not miss it, I just had no time yet to respond. I wanted to respond to Mod.Sense first to ease his frustration ;-)
Target CPU runtime is new to me. Does this actually mean the size of the WU I receive? I assume setting a Target CPU Runtime affects all computers? Also, why should I set a smaller cache size than the default 0,1/0,5 and even smaller if I also run another project? The system settings on limiting BOINC when the computer is in use and the time available I have determined after having found the system not responsive enough while using it when tasks were still running and the temperature rising too much to my liking. I started with the defaults and then scaled back when I stumbled on some limitation and thus arrived at my current settings where I do not have any issues anymore. I'd rather keep it that way.

BOINC allows multiple projects to be run on a single (or multiple) systems. The default settings will work for pretty much any project & system, they are set for the vast majority of users & projects. However for edge cases such as yours, and when more than one project is being done, changes to the defaults will be required to meet those particular circumstances.

Yes, I think you're right, I did not realize that.

In your Account, Computers, Details. Also per application- Application details, Show.

There it is! I did look for them before posting but could not find it, but I did remember that I saw them sometime. With "per application- Application details, Show." do you mean the BOINC client or somewhere else? In the BOINC client I only see Project > Properties and then no average turnaround time.

From your earlier post:

The default target CPU Runtime is 8 hours. To do 8 hours work over 3 days is not a big ask or expectation.

I disagree here. On my weaker computers this apparently is an issue. World Community Grid has longer deadlines for smaller tasks, so a longer deadline can be set. I am still curious why the deadline had been reduced from 7 to 3 days. Before that happened everything was running fine with the default settings on all 3 computers.

Also, why should I set a smaller cache size than the default 0,1/0,5 and even smaller if I also run another project?

The idea of a cache was from the days when people didn't have 24/7 internet access, and many projects were less than stable.
Given how stable Rosetta is, and if running more than one project the likely hood of them all being down at the same time is minimal, and even with different data pricing at different times of the day, you're better off with the smallest possible cache.
The smaller the cache, the more time a system is able to process work while it is on, then the sooner your Resource share settings can be met, and the less likely deadline issues will be.

Part of the problem is also the Estimated completion times tend to be less than the actual CPU time (which is always less than the Runtime). And when new applications come out, they tended to be a huge amount less. The project recently made changes so that shouldn't occur again- when it sends out work it no longer uses the Estimated completion time to determine how long it takes to process a Task, but uses the actual Target CPU time instead. So it should reduce the number of people having problems with deadlines.
But even so, having a small cache will still help in that regard- particularly if people have "Use at most 100 % of CPU time" at anything less than 100% Setting that to 50% means it will take a minimum of 16 hours to process an 8 hour Target CPU runtime Task. Hence it's best to leave that at 100%.

do you mean the BOINC client or somewhere else? In the BOINC client I only see Project > Properties and then no average turnaround time.

On your Account page, part of the Computer details, Just below "Average turnaround time" is "Application details", click on Show.
There it will have the Average turnaround time for each of the applications the system has used to process work.

I am still curious why the deadline had been reduced from 7 to 3 days. Before that happened everything was running fine with the default settings on all 3 computers.

Many of the present Tasks are for work being done on COIVID-19 (either directly or indirectly). Given the World situation with the virus, the sooner a vaccine is developed, the better. Hence the shorter deadlines- much of the work is time critical and sooner is better.

---

Grant
Darwin NT

Target CPU runtime is new to me. Does this actually mean the size of the WU I receive? I assume setting a Target CPU Runtime affects all computers?

You can set it for Rosetta only and will apply to all 3 of your computers.
In the Boinc manager, select Rosetta and on the left, click Your Account
In the webpage it sends you to, chose Rosetta@home Preferences and edit your Target CPU runtime.
It sounds like 4hrs will suit you and allow tasks to be returned before deadline.
Hopefully it works for you.

---

Is that true? lately some tasks (maybe one in 10, or one in 15) come in with reasonable estimated times, and as they run the time remaining keeps increasing! If I don't abort them, they keep running. And I start missing deadlines on ones that haven't even started yet.

I just aborted one which started off estimated 8 hours. It had been running for 20 hours, and the estimated time remaining was 35 hours? What do we do about that?

No, the other solution that i pointed out & you seem to have missed is that you can set the Target CPU Runtime to a value that is less than the 8 hour default, and set a very small cache size (with more than one project, no cache is best). And you can change the settings on your system so Rosetta is able to do more work while the system is running without impacting on it's general performance/responsiveness.

I read that but can't find anywhere to set Target CPU Runtime. I'm lost. I've been running BOINC for, gee, a lot of years, and Rosetta for a few. The last couple of months I've been running only Rosetta tasks, and I've added a much faster computer. But I feel bad about having to abort tasks, or having them miss deadline because they are waitig to start.

Edited to add: I found it. Set it to 6 hours. Min queue set to .3, amx additional to .7. Let's see hw this works.

I read that but can't find anywhere to set Target CPU Runtime.

Your account, Preferences, Preferences for this project- Rosetta@home preferences, near the bottom is Target CPU run time.


You'd be better of keeping the default Target CPU runtime, and figuring out what is using all the CPU time on your AMD system- almost half if it's CPU time is spent doing things other than processing Rosetta work.
One of your Valid tasks-

    Name r4k_21113_fold_SAVE_ALL_OUT_919395_1108_0
Run time 14 hours  5 min 13 sec
CPU time  7 hours 55 min 43 sec

On a lightly used system,. the difference between CPU time & Runtime for an 8 hour Task, should be around 4min. So something else is using a lot of CPU time.

---

Grant
Darwin NT

No, one machine (the older an slower one) is doing nothing else for about 12 hours a day, and usually over 50% (often waaaaay over) on Rosetta tasks for 12. The faster computer is doing nothing else 24 hours a day.

You'd be better of keeping the default Target CPU runtime, and figuring out what is using all the CPU time on your AMD system- almost half if it's CPU time is spent doing things other than processing Rosetta work.
One of your Valid tasks-

    Name r4k_21113_fold_SAVE_ALL_OUT_919395_1108_0
Run time 14 hours  5 min 13 sec
CPU time  7 hours 55 min 43 sec

On a lightly used system,. the difference between CPU time & Runtime for an 8 hour Task, should be around 4min. So something else is using a lot of CPU time.

Multiple Rosetta tasks are running concurrently. If that's not what should be happening, where can I change that?

Multiple Rosetta tasks are running concurrently. If that's not what should be happening, where can I change that?

The problem isn't multiple Tasks running, the problem is your system is doing lots of other CPU intensive work other than Rosetta, that is why there is a huge difference between Run time & CPU time.
Rosetta Tasks run at idle priority, so pretty much any other programme running on your system will get processing priority over Rosetta, and that is what is happening.

No, one machine (the older an slower one) is doing nothing else for about 12 hours a day, and usually over 50% (often waaaaay over) on Rosetta tasks for 12.

What are you doing with that machine for the other 12 hours? Whatever it is it is very CPU intensive.


From memory Task Manager on Win8.1 isn't much better than that on XP/Vista. I'd suggest using Process Explorer to see what else is using your CPU time so heavily.

Process Explorer https://docs.microsoft.com/en-us/sysinternals/downloads/process-explorer

---

Grant
Darwin NT

On the website, on your Account page, in Preferences for this project, you can also set different preferences for different Locations then change your computers' Locations appropriately. The physical Location doesn't matter; they could be sitting side by side, as mine are.
You could set a faster machine to run 8hr tasks as Primary (Default) and a lesser machine to run 2hr tasks as Home and perhaps another with different preferences again as School.

And in Boinc Options/Computing Preferences you can also set the % of the CPUs to use eg. 75% for 3 0f 4 cores, 50% for 2 of 4, or 66.67% for 16 of 24, if you're lucky enough to have that capacity, so that Boinc uses some cores and some are always available for exclusive use of the OS and your other uses.