Too short deadlines

[jsantix]

Hello there.

Currently I'm folding on a company laptop, which I only use around 8 hours a day, M-F.
I've configured it to use just 1 CPU thread in order to not disturb my job.

The problem is, the WU I'm being assigned are usually just 1 day deadline, but it takes me about 2-3 days to complete them.
I usually finish them before the expiration date, so that is not the problem, but I would prefer if I'm assigned WUs with longer deadlines, so the same WU has not to be always sent to other machine, as I feel like my folding time is not being useful at all.

Is there a way to configure my client to ask for such long-deadline WUs?

Another option I can consider is to increment my CPU threads in order to complete each WU in a single day, but in that case I would like to ensure I am not assigned a new WU until the next working day, as it would be worthless if I am assigned a new WU on Friday evening and I cannot complete it until next Monday.

I've read about max-units flags but it will just pause my work when a WU is finished, but I would have to resume it manually the next day.
And I'm not even sure if I will receive a new WU just before the client is paused...

Can you help me to choose the best client configuration for me?
Thank you in advance!

[Jonazz]

Hello and welcome!

If you constantly miss the timeout deadline, then I'm afraid your laptop is too slow to fold.

As for your other option, you can set your client to "finish". This way the FAH client won't download a new WU when you're finished with the current one.

[jsantix]

Thanks! That would be great. How can I do that?

And yes, it is slow... But do you think it is totally useless, or as long as it finish the tasks before the expiration date, it could be somehow useful?

[Jonazz]

Open the FAH client (not the browser). Somewhere in the top left you will see the "Fold", "Pauze" and "Finish" buttons.

After the timeout expiration, your WU is also sent to someone else. So as far as I know it's actually harmful

[ajm]

As of now, it is not possible to ask for a particular deadline. The server will just match your hardware with the work units, but that's about it. If you don't return a WU in time, it will be assigned to someone else. If you nevertheless finish it, the system will give you some points for that, but the job will actually have been done for nothing, as it will have been assigned to another machine.
You should be able to allocated more than one thread to FAH, though - these jobs run at the lowest priority, so that they should not disturb you (apart from the fans's noise and the heat, depending on your hardware).

For the Finish option, it is in Advanced Control (aka FAHControl). In Windows, you can open it in the taskbar, on the right side, under ^. You'll see there the typical FAH icon, click it and choose Advanced Control.

[bruce]

I've configured it to use just 1 CPU thread in order to not disturb my job.

The FINISH setting (above) is an important addition, but you also should reconfigure the number of threads. The CPU processing done by FAHCore_a7 is done at a very low priority and it should back-off processing when foreground processes for you work have any work to do.

[Jonazz]

As of now, it is not possible to ask for a particular deadline. The server will just match your hardware with the work units, but that's about it. If you don't return a WU in time, it will be assigned to someone else. If you nevertheless finish it, the system will give you some points for that, but the job will actually have been done for nothing, as it will have been assigned to another machine.

I think OP is not the only one in this situation. I'm sure there are more people believing that they're contributing to the science by meeting the timeout deadline, while they're actually slowing things down (nothing personal OP). Maybe the difference between both deadlines should be explained more thouroughly to new folders.

[Neil-B]

... or even to those who profess to understand them

FaH has two defined time periods ... Timeout is the point at which the researchers have chosen to allow for reissue of a WU in case the original one has been lost or delayed - this does not necessarily happens immediately as the WU is put in the queue ... Expiration is the point at which the researchers have chosen that a WU no longer has value and it is dumped ... Anyone completing a WU within Timeout is contributing to the progress of the project - fact ... Anyone completing a WU within Expiration and before anyone else has completed it is also contributing to the progress of the project - also fact ... and as clarified by Joe_H later in this thread anyone completing a WU within Expiration irrespective of whether anyone else returns it is still contributing something.

Now, as to whether people participating with slower or non-uber gear is slowing down the progress of projects depends on how you look at the FaH capability as a whole ... Under traditional conditions where work exceeded the resource available and all resource was busy then without doubt the presence of slower kit got more WUs done and overall projects progressed faster - 5 day Timeouts were pretty normal ... at the moment where at least for a few months there has been more resource than work then a case can be made that slower kit might be impeding progress of WUs - no doubt one of the factors in Timeouts shortening dramatically is that with excess resource makes it possible to reissue early and potentially double up successful folding of a WU with impacting overall progress and potentially speeding up some project flows ... imo this is less to do with slower kit holding up progress and more to do with a luxury of options ... However as soon as there is enough work to keep all resource busy the the idea that the slower kit is holding up progress overall is a fallacy - the faster kit just does other stuff faster so if you remove the slower kit FaH capability and overall throughput actually suffers.

Now there are projects that the Researchers want done asap and the are prioritised to some extent by species or core count and by shifting Timeouts to reissue very early ... the presence of slower kit will impact here but that is why restrictions and early Timeouts are set ... but a bigger impact is more likely folding patterns - faster kit is slower than slow kit if only run intermittently around gaming or lifestyle patterns and a fair bit of the faster kit may well need to be throttled back or shut down for periods due to hot weather.

As resource continues to drop for the many varied reasons and the work begins to once again significantly exceed this then actually without the slower kit folding FaH would suffer overall once more ... Whilst there are riches it is easy to be picky ... When there is a drought one takes everything one can get

[bruce]

And with that explanation, the timeout should be long enough that a significant percentage of hardware platforms should be able to complete the WUs assigned to them within that time provided they run 24x7. Setting Timeout values so that too many folks fail to meet them causes FAH to flood the environment with excess WU duplications.

[RaindropsKeepFalling]

Just came here to ask similar things. So is there no way I can simply abandon a work unit, so it can be released to someone else? I've completed two work units so far, so my old machine is perhaps not completely useless, but the current WU has an ETA of 4.32 days while the deadline is 3.61 days, so I'm tying up my machine with pointless processing, and delaying the project accordingly.

[Jonazz]

Abandoning a WU hurts the science, so that's not a good strategy.

[Neil-B]

For a WU that (having run for at least 5/10% is estimated to complete outside the Expiration there is no point in continuing to fold the WU as it just won't be returned in time to be of value ... and if a slot is having this issue on a significant number of WUs it attempts then it is simply too slow and every attempt to fold a WU is just putting a delay equal to at least the Timeout on the WUs in question and so the slot should probably be retired ... if however the slot can be "improved" by say adding another few threads/cores or by running for more hours/more regularly then there may still be a potential use for it

Under traditional conditions where there is lots of work and limited resource as long as the WUs are completing within Expiration there is a reasonable chance that the WU will still help FaH progress by being the first completed as any reissue may not happen straight away at Timeout however it is true that under current circumstances where there is still quite a lot of resource it appears that reissued WUs are picked up fairly soon after Timeout ... Keeping an eye on the WU status https://apps.foldingathome.org/wu will show if anyone has returned the WU completed "OK", at which point there is little/no direct scientific benefit to FaH in returning the WU again (although researchers do sometimes use double returns to check validity of results and other factors) - however it will still get points as long as it is returned before Expiration ... The choice of whether to dump the WU at this stage is a personal call - leave it running and gain some points for the work done even if of little value at this point to FaH (if one is concerned about points) or dump it and start working on a new WU that hopefully will help FaH progress.

It is a kind of hard call - up until the point another folder returns the WU successfully continuing to fold a WU that may complete after Timeout but within the Expiration is imo the right thing to do as it is potentially adding value as it may still be the first to complete the WU ... If a slot that is never (or regularly fails to be) the first to complete a WU but can complete within Expiration then it might be worth pausing the slot for a while until the resource levels abate (days/weeks/months) occasionally running a WU just to see if resource has dissipated to the point that it can actually return WUs before others and help FaH progress ... However if one is happy to keep running the slot it will still be giving researchers the opportunity to cross check results even if not being the WU return that allows the next generation to be released.

For example I occasionally un a 2/4 core slot when there is an abundance of small WUs - currently completing 13415s in sub 5 hour so is adding value and helping this project progress - but I do keep a close watch on it - it isn't in danger of exceeding Expirations but if certain big WUs appear it struggles to complete within Timeout so I start monitoring it to see it is still returning the WUs first - if not I pause it until my other kit shows me small WUs are back in favour.

These are "strange times" at the moment - more resource than work is unusual to say the least ... The FaH Team and the researchers have massively expanded throughput of projects and are developing new wonderful ways to run projects to utilise the resource but there are still "lulls" in throughput ... Work may well also being done behind the scene to help assignment and optimise to some extent WU allocation but this is non trivial to develop and simply takes time (especially since breaking what FaH has already got is not really an option ... with the exception that there are a few projects around at the moment that are specifically pushing new boundaries/limits and where WU failure is actually in many ways a "good result" as it is helping the Researchers tune these new methods at real pace - but again this is not the usual state of play !!

[jsantix]

Thank you all for your responses.

It's sad but if I'm not actually contributing (or even worse, delaying the project), I think I'll stop folding for a while, at least after summer ends, as this laptop gets too hot while folding even with a single core.

Cheers!

[bruce]

Thank you all for your responses.

It's sad but if I'm not actually contributing (or even worse, delaying the project), I think I'll stop folding for a while, at least after summer ends, as this laptop gets too hot while folding even with a single core.

Cheers!

Maybe your configuration can be tweaked to help that situation, We can make suggestions. ... but we'd have to see the first couple hundred lines of FAH's log so we understand your configuration. See the sig block of my first post above.

[jsantix]

Maybe your configuration can be tweaked to help that situation, We can make suggestions. ... but we'd have to see the first couple hundred lines of FAH's log so we understand your configuration. See the sig block of my first post above.

Better than logs, take a look at this screenshot:

https://i.ibb.co/TRVRmRz/Captura.png

As you can see, I was assigned that WU on 26th, and despite the expiration time is 1 week, the timeout is just 1 day.
If you take a look to the ETA, with already 42% done, it will still take > 6 hours to finish, so the total WU would need around 10~12 hours to complete in my laptop, but I only use it while working, around 8 hours a day.

And if you check this: https://apps.foldingathome.org/wu#proje ... e=1&gen=78
This WU was assigned to another client just 15 minutes after the timeout deadline, and it was completed in just 5 hours.

So unless I manage to complete my WUs in less than 8 hours, my contribution is just slowing down the project, which is very disapointing...

I would like to assign a second thread to FAH, but it is too hot here and it would probably set my laptop on fire!!
If they could just increase the timeout to 2 days instead of only 1...