Is not completing WUs on time useless ?

[Krouiksi]

I'm thinking about advertising F@h to my parents. They have an old-ish computer that only runs a a couple hours a day.
They probably would not be able to complete WUs by due date.

Will their donation be useless to F@h, or will this only prevent them for gaining points ?

I couldn't find an answer to this in earlier topics, I'm sorry if it already exists.

Thank you for your hard work, especially nowadays !

[JimboPalmer]

If they cannot complete by the time out, it is reassigned to others and your contribution is useless.

It really is not for everyone.

[bruce]

Yes, projects which cannot be complete by the expiration date are useless. (in fact, they are a slight negative value)

Sorry, but a couple of CPU hours per day can't help the research.

[Krouiksi]

I see! That was my guess, but I needed to be sure.

Indeed there's no point in using your servers & delaying the projects because of kind but useless "help".

Thanks

[Frogging101]

Yes, projects which cannot be complete by the expiration date are useless. (in fact, they are a slight negative value)

(Emphasis added)

Do you mean the timeout date? My understanding is that a WU is "reserved" for a client (it won't be assigned to any other client) until the Timeout. If it isn't completed by then, it will be reassigned. If the first client returns it after that, then it will just be duplicate work and not useful to the project (though it may still be credited some points). And the server won't even accept it at all after the Expiration.

Is it ever useful (for science, not for points) to return a WU after the Timeout? Maybe in the rare case where the client it was reassigned to ultimately fails to complete it and the late result is the only good one?

[PantherX]

...My understanding is that a WU is "reserved" for a client (it won't be assigned to any other client) until the Timeout. If it isn't completed by then, it will be reassigned. If the first client returns it after that, then it will just be duplicate work and not useful to the project (though it may still be credited some points). And the server won't even accept it at all after the Expiration.

Is it ever useful (for science, not for points) to return a WU after the Timeout? Maybe in the rare case where the client it was reassigned to ultimately fails to complete it and the late result is the only good one?

Under normal conditions that's true but there are some other edge cases too.

If a client returns a WU after the timeout and before the expiration date to the Server and it's the first one to return that WU, the science will progress forward without issues and that work is very useful. Do note that the second assignment of the WU will be placed at the "back of the line" for it to be assigned, it doesn't "jump the line" just because the timeout period was reached.

Also, if the WU has reached the expiration date, the client will dump it, it will not bother to upload it to the server since it is a waste of resources.

[Rel25917]

For the science i think they can compare duplicate work and make sure it matches so still some use. Usually the timeouts were much closer to the final deadline but they want covid stuff done fast so the timeout is much shorter.

[The Real Weasle]

Thank you for the explanation of the difference between TIMEOUT and EXPIRATION.

I had not actually noticed the difference, I've just glanced at the Web Control that says that my computer has 7 days to complete the WU or whatever which is the Expiration.
Doesn't affect me or Folding@Home as I have 3 machines on 24/7 - at least until the temperatures pick up, but good to know. Even the slowest machine completes well within the Timeout.

[peterjammo]

Has there been a stepchange in short timeout wu's in the past few days? I've only been folding for 3 weeks, but for the first 2 it was rare that my old core2duo laptops couldn't complete within timeout. In the past couple of days I've picked up at least 3 wu that my slower machines can't do inside timeout, and 2 that my faster core2duo can't make. I'm waiting for an upgrade cpu for my faster laptop which should hopefully bring it back inside the envelope, but I don't have that option for my 3 older machines as they don't support the faster fsb c2d cpu's.

It seems a shame to switch them off since they've been returning at least a wu a day on average each, usually well inside timeout, but if the current trend is to shorter timeouts than they can manage, it would probably be the sensible thing to do. Sods law being what it is, the slower machines have picked up all the large tight timeout wu for the past few days, while my 4 core box has either sat idle or been churning through small wu with several day timeouts.

Any info on timeout policy change?

[Neil-B]

Many of the COVID-19 related are 1day Timeout https://apps.foldingathome.org/psummary for CPU work ... there are longer though - it may be luck of the draw ... There are some really tight Deadlines appearing for some of the Projects which I guess reflect the urgency placed on those projects.

Outside Timeout but within Deadline (or Expiry Date) is still fine - someone else may get issues it but if completed within deadline it will still be uploaded and has value.

[FoldingFodder]

Do note that the second assignment of the WU will be placed at the "back of the line" for it to be assigned, it doesn't "jump the line" just because the timeout period was reached.

What do you exactly mean by placed at the back of the line?

Has there been a stepchange in short timeout wu's in the past few days? I've only been folding for 3 weeks, but for the first 2 it was rare that my old core2duo laptops couldn't complete within timeout. In the past couple of days I've picked up at least 3 wu that my slower machines can't do inside timeout, and 2 that my faster core2duo can't make. I'm waiting for an upgrade cpu for my faster laptop which should hopefully bring it back inside the envelope, but I don't have that option for my 3 older machines as they don't support the faster fsb c2d cpu's.

It seems a shame to switch them off since they've been returning at least a wu a day on average each, usually well inside timeout, but if the current trend is to shorter timeouts than they can manage, it would probably be the sensible thing to do. Sods law being what it is, the slower machines have picked up all the large tight timeout wu for the past few days, while my 4 core box has either sat idle or been churning through small wu with several day timeouts.

Any info on timeout policy change?

I have a decent Quadro GPU in a 2 year old PC (built to handle complex 3D models with ease) that's on 24/7 and that missed the time out by 3 hours today - the computer hasn't been used for the past couple of days, so the only thing it's done is process F@H. So it's not just weak and ancient computers that are struggling to meet the deadlines.


I think the time out deadline needs to be relaxed a little and/or WUs be better targetted at the appropriate hardware. I imagine the number of WUs returned after the timeout deadline has increased this past week.

[Neil-B]

We need to be careful/clear about Timeout and Expiration (or Deadline) are two different things … Timeout is when the WU gets put back into the queue for reissue and after which time WU only accrues Base Points … Expiration (or Deadline) is when the WU will no longer be accepted back by the WS/CS (maybe even dumped by client iirc) and has no points or scientific value.

[FoldingFodder]

We need to be careful/clear about Timeout and Expiration (or Deadline) are two different things … Timeout is when the WU gets put back into the queue for reissue and after which time WU only accrues Base Points … Expiration (or Deadline) is when the WU will no longer be accepted back by the WS/CS (maybe even dumped by client iirc) and has no points or scientific value.

For me, the timeout is the deadline; otherwise i might as well have saved myself some electricity and not processed the WU since another machine is processing the WU as well. I'm a person who doesn't care about worthless internet points, just like Reddit karma and Facebook likes.

[Neil-B]

... Just wanted some people who might read this thread to know there is still scientific value in completing after timeout but before expiration ... it is still perfectly possible to be the first person to return a WU even once timeout has passed - and that is good for the science even if the points are only base

[PantherX]

...What do you exactly mean by placed at the back of the line?...

Here's my (simplified) understanding

Under normal conditions (demand = supply) here's the current waiting list for new WUs:
New WU A -> Assigned
New WU B -> Assigned
New WU C -> Assigned

Let's say that WU B and C returned and this is what happens on the Server:
New WU A -> TImeout
New WU B -> Returned
New WU C -> Returned

The Server queue is now:
New WU B2 -> Assigned
New WU C2 -> Assigned
New WU A -> Re-assigned

Of course, there have been several tweaks and changes over the years to the F@H Server code so things may not be what they used to be.