Hello,

I lost results during the long outage. I know that I lost these results, because they were the fourth results of already validated WUs and are not needed anymore in scientific terms.

But why do you ignore these not needed but valid results?

SETI sent out the work to the crunchers. The crunchers have done the work. The crunchers got valid results. The crunchers finished the results in time. And even the crunchers of the fourth results can at least expect on thing: To get the credit for the work they have done. And after a long outage they still must have the chance to get the credit for it. If you cannot guarantee this, the system is wrong and should be changed.

Most likely, the crunchers of the fourth results are always the same, because they have slow computers, do not run their computers all the time, and so on. On one hand you profit from sending out four results because you get valid results faster. On the other hand you punish the crunchers of the fourth results for beeing the last by ignoring their valid results during a long outage, although they didn't even got any chance to upload their results.

With this attitude you degrade the 4th-result-crunchers to second class crunchers. And this is not acceptable.

Here are some suggestions:

Mark every WU with results still in progress and still before the deadline with a "don't touch me"-marker if an outage has started. Give the crunchers of *all four* results enough time to upload the results after the outage.

Similar to the first suggestion: Automatically expand the deadline during an outage. Each day of an outage shifts the deadline for one day.

These two suggestions probably mean a total redesign of the software.

Therefore most likely the simpliest suggestion: You need three valid results? Send out three results. Don't take the advantages of sending out four results but ignore one result if it is not needed anymore.

Kind regards,
Carsten

If this is truely the way it is handled, then I can see where you could lose crunchers. Imagine the guy with the 3.xGHz machine that is on dial up. connects once a week, crunches umpteen results, faithfully connects once or twice a week. But because the other three got theirs in on day two, and he connects on day three, his umpteen results mean nothing, and are rejected. Not a good scenario.
Same for anyone that faithfully crunches, connects, even the slower machines on 100% of the time 100% online 100% crunching. They will get their results in after getting one after only a day, and the speedy machines that got theirs in within 2-6 hours of getting them, get the credit, and the slower one doesn't. Continually, then, to make matters worse, he has more to go, that are probably already validated by the three others before he can even run them.
I know, not a perfect world, and the sending 4 method does ensure that the results are validated quicker, due to probably 10% of the time one of the 3 doesn't respond, or sends bad data, or????
OH well, catch 22, what do you do.
Either way, someone is screwed.
Either way, someone will yell.

I do vote for the send 3 to get 3, and if after 1/4 of the deadline you don't have all results, send it again to a 4th. Since most serious setiers will get theirs back in a few days, usually 1/4 the deadline, this should work.
Yep, rewrite, and also, .... sorry, rambling...

Patience, Persistence, Truth

---

Thanks, and Keep on crunchin'
John Henry KI4JPL
Sevierville TN

I started with nothing,
and I still have some of it left.
<img src="http://www.boincstats.com/stats/banner.php?cpid=989478996ebd8eadba8f0809051cdde2">

Hello,

I lost results during the long outage. I know that I lost these results, because they were the fourth results of already validated WUs and are not needed anymore in scientific terms.

But why do you ignore these not needed but valid results?



Kind regards,
Carsten

Ordinarily, there is no penalty involved with being the fourth result in a work unit. As long as they come in before their deadline they get credit regardless whether the WU validated on three results or not.

The problem his time was the reason for the long outage was the need to purge the file system of the antique results which had accumulated, and an unfortunate side effect was that some slower returning "fourths" got hung out to dry so to speak during the validation/purging process. However, this was preferable to the quick and dirty approach which would have been to merely delete everything and start from scratch.

Alinator

Just one last word. (yeah right)
This 4 out, first 3 back = good, 4th one back = no credit, will lead to a lot more people panicing when the outages occur.
People will be more prone to sit there and continually hit retry until theirs are in, for fear of losing credits because they weren't one of the ones continually hitting retry and ended up being the 4th one in.

I vote for an even different way.
You want 3 valid similar results to post credit.
Fine, when the first three come in, compute credit, validate, and post their credit.
As long as the 4th is within the deadline (as extended by outage days), and it is validated similar to the other three, then the 4th should also get his/her rated credit.

sorry, a new argument will brew, time will stop because we think differently,
OH NO MR. BILLLLLLL.
sorry, digressed.
Think about it, there may be ways to alter a little code to help maintain some peoples desires.

---

Thanks, and Keep on crunchin'
John Henry KI4JPL
Sevierville TN

I started with nothing,
and I still have some of it left.
<img src="http://www.boincstats.com/stats/banner.php?cpid=989478996ebd8eadba8f0809051cdde2">

ALinator, I was typing when you were, and if I had read your message, then my suggestion would not have occurred. Thanks for the input. I do think then that the only addition needed would be to extend deadlines by days down. This seems easy, but may be nasty as deep doo doo if the code isn't written for adjustable deadlines.
Thanks,

Hello,

I lost results during the long outage. I know that I lost these results, because they were the fourth results of already validated WUs and are not needed anymore in scientific terms.

But why do you ignore these not needed but valid results?



Kind regards,
Carsten

Ordinarily, there is no penalty involved with being the fourth result in a work unit. As long as they come in before their deadline they get credit regardless whether the WU validated on three results or not.

The problem his time was the reason for the long outage was the need to purge the file system of the antique results which had accumulated, and an unfortunate side effect was that some slower returning "fourths" got hung out to dry so to speak during the validation/purging process. However, this was preferable to the quick and dirty approach which would have been to merely delete everything and start from scratch.

Alinator

---

Thanks, and Keep on crunchin'
John Henry KI4JPL
Sevierville TN

I started with nothing,
and I still have some of it left.
<img src="http://www.boincstats.com/stats/banner.php?cpid=989478996ebd8eadba8f0809051cdde2">

In the tech news they stated that the results that came back late will get credit, if valid.

Currently some are getting 0, this happened before. They have a program they will run at a later point (after all the hustle and bustle is over probably) that will grant those credits.

Patience is all that is needed.

---

My movie https://vimeo.com/manage/videos/502242

I am guessing that the reason fourth work units are generated is because it is easier to generate four copies at once rather than having to go to the database and generate additional work units for times when one or more results are returned with errors.

In the tech news they stated that the results that came back late will get credit, if valid.

Currently some are getting 0, this happened before. They have a program they will run at a later point (after all the hustle and bustle is over probably) that will grant those credits.

Patience is all that is needed.

I had a few results caught in this scenario.

The case was if there were three results returned for a WU before the outage occured.

During the file system purge those WU's validated on three, marked mine as "Didn't Need" and then got purged.

For all WU's I had that had less than three results returned, I have gotten credit for when uploaded.

In any event, there were only two or three them, so I'm not losing any sleep over the "lost" credits. ;-)

Alinator

I also have posted, after the most recent outage, several "fourth" (class), results to get ZERO CREDIT!!!!
It's pretty frustrating... We can holler as much as we want, but nothing will be done...

---

LarryB56

I also have posted, after the most recent outage, several "fourth" (class), results to get ZERO CREDIT!!!!
It's pretty frustrating... We can holler as much as we want, but nothing will be done...

For those with zero credit did the WU validate OK or not? Have a look at these WUs, then click on the result ID link and post back what is says under the "stderr out" and "validate state" sections. Maybe there are some clues there to your problem.

---

The 4th one always gets credit, occasionally it takes a little longer for the 4th but they get it.

The antique deleters did not tuch the WU's waiting for credit or waiting for the 4th to come on for credit as someone above said they were deleted. Read other threads on what was going on and you will see that they were way to new for what the antique deleter was doing. Also check your results page, still goes beck to early August.

Only one copy of made of each WU, and that one copy is sent to 4 diferant computers. Not 4 copies made like said above. Just think about it, if 5 people want copies of a file that you have on your computer do you make 5 more compies before sending them a copy in the email? Or just attach your file and a copy is sent?
Ray

---

Pizza@Home Rays Place Rays place Forums

So the 4/3 WU system has started in the beginning of this year,
because people have to wait a very long time for their credits.

Before the 4/3 systems started there was the 3/3 system. So if
one result didn't come back, the fourth copy was send after the 14 days
delay, and if this result either didn't come back you have to wait again
14 days till it was validated.

So they decide to start the 4/3 system where (normally) after 3
revieved valid result credit is granted and when the 4th result returns
in time it will get the calculated credit too.

if 4 results return before the validator grants the credit the
credit will be calculated by the average of the 2 midst claimed results.

Right now the db_purge and file_deleter are disabled so no result is
out of time. if a forth result now get 0 credit, something in the calculation
went wrong and the result is not valid. Doesn't have something to do
with the outage.

So the average 'active' time of a WU is reduced and the credits grant faster.

and - because of the limitation of the recorded data - there is more
WUs which have to be calculated. So they were sent 4 times not 3 times and not
infinite like in classic seti, just keeping the people happy by giving
them something to work.

---

Hello,

In the tech news they stated that the results that came back late will get credit, if valid.

You are right. But the tech news also states that the canonical results still have to be on disk. And if the results were marked as "not needed" before the deleters were turned off I don't see any chance to get credits for these results.

Currently some are getting 0, this happened before. They have a program they will run at a later point (after all the hustle and bustle is over probably) that will grant those credits.

I hope you are right. But how can you validate a result if its siblings have already disappeared?

For those with zero credit did the WU validate OK or not? Have a look at these WUs, then click on the result ID link and post back what is says under the "stderr out" and "validate state" sections.

The validators didn't even have a look at these results. They got "granted credit: 0.00" immediatly after the report.

Result 100285890
Result 100285894
and others

The 4th one always gets credit, occasionally it takes a little longer for the 4th but they get it.

But than the WU page should state in the "granted credit" column "pending" and not "0.00". Or am I wrong?

Carsten

I also have posted, after the most recent outage, several "fourth" (class), results to get ZERO CREDIT!!!!
It's pretty frustrating... We can holler as much as we want, but nothing will be done...

For those with zero credit did the WU validate OK or not? Have a look at these WUs, then click on the result ID link and post back what is says under the "stderr out" and "validate state" sections. Maybe there are some clues there to your problem.

Spot on Dude!!! The Validate State read results invalid!
Thank You.

---

LarryB56

It has been argued in other threads, by people with fast hosts and low benchmark scores, that being the fourth result to return boosts their credits score. You'll either have to find the threads yourself or do the maths.

I'm sure most of them are not producing bad science but feel they want just returns for their efforts.

Andy

Hello,

In the tech news they stated that the results that came back late will get credit, if valid.

You are right. But the tech news also states that the canonical results still have to be on disk. And if the results were marked as "not needed" before the deleters were turned off I don't see any chance to get credits for these results.

Currently some are getting 0, this happened before. They have a program they will run at a later point (after all the hustle and bustle is over probably) that will grant those credits.

I hope you are right. But how can you validate a result if its siblings have already disappeared?

For those with zero credit did the WU validate OK or not? Have a look at these WUs, then click on the result ID link and post back what is says under the "stderr out" and "validate state" sections.

The validators didn't even have a look at these results. They got "granted credit: 0.00" immediatly after the report.

Result 100285890
Result 100285894
and others

The 4th one always gets credit, occasionally it takes a little longer for the 4th but they get it.

But than the WU page should state in the "granted credit" column "pending" and not "0.00". Or am I wrong?

Carsten

Thanks for posting the links to the WU's that show this effect. In my second post I incorrectly said that mine got marked as "Didn't Need", when in fact the results in question were exactly like the Zero Credit examples you pointed out. That's what happens when you don't look at the logbook before posting! ;-)

Please note that the reason for this is because the WU validated on the three results which came in before the outage, and the deadline for zeroed result was *during* the outage, and therefore *couldn't* be reported regardless of whether there was a client error or not. In your case (as in mine) there is no indication that your result was faulty in any way, other than it was late.

In your case you may still get credit from a backside correction since your result are still shown in the "active" results. I doubt I can get credit for mine since as I said before they got purged during the FS cleanup and no longer appear in the "active" results.

I'm hypothesizing the reason for this is that early on in the outage UCB left the normal file handling processes running as well as the antique deleter in order to speed things up as well as get a better understanding what the problem was with the file system overall and develop better stratgies for addressing it. At some point they must have noticed this effect and realized it would result in a small percentage of valid result not getting credit for work done as well as contribute to new orphans being created when those completed results came in after the project went back online. This probably explains why they haven't fired up the normal file deleters and the assimilators yet, since they want to make sure that everything that can come in has come in.

Alinator

a short update

Pooh Bear 27 wrote:

In the tech news they stated that the results that came back late will get credit, if valid.

Currently some are getting 0, this happened before. They have a program they will run at a later point (after all the hustle and bustle is over probably) that will grant those credits.

Patience is all that is needed.

Well, patience didn't help...

Finally, these 0-credit-results disappeared from my result pages without getting any credit. As far as I understand the system the WU has now been removed from the database. So, there is no chance anymore to get any credit for these results.

to summarize at least for my own results:

Is there a program that "repairs" the "0-credit-during-outage"-results? No.
Do 4th results always get credit as Ray Brown stated? No.
Did I loose work just because I wasn't able to upload the results in time during the long outage? Yes.

Please draw your own conclusions if the statement in the topic of this thread is right or wrong...

Carsten

Well, Carsten, I agree with you on one point you made in your original post in this thread. I believe it would be best if the '4th initial result' were not issued, and S@H goes back to their original design of only issuing 3 initial results. The reason, as you said, for the 'initial 4th' result was to speed granting on credits (so that one late return would not hold up credits for everyone else on that work unit). I have never agreed with this decision, but then, you can't please everyone. I can wait a while for my credits, since credits are only a secondary reason I participate.

One benefit to the project of dropping the issuance of a 'initial 4th result' would be to decrease the number of files in the download directory by about 25%, assuming the number of issued work units is constant. This would help with the current problem of 'too many files' in the download directories, and give the project a little more room to grow on the current disk hardware.

---

https://youtu.be/iY57ErBkFFE

#Texit

Don't blame me, I voted for Johnson(L) in 2016.

Truth is dangerous... especially when it challenges those in power.

One benefit to the project of dropping the issuance of a 'initial 4th result' would be to decrease the number of files in the download directory by about 25%, assuming the number of issued work units is constant. This would help with the current problem of 'too many files' in the download directories, and give the project a little more room to grow on the current disk hardware.

Actually, I think it would have the opposite affect. Instead of decreasing the number of files, it would actually increase the number of files. The problem they found was that a large percentage of results were not being returned, therefore the WU's had to be reissued which added another 2 week delay. now we have those files on the RAID for 4 weeks (and sometimes more), whereas if they issue 4 initially, but only 3 get returned, they still get a quorum and the files can be deleted after the initial 2 weeks.

So, by speeding things up, they can get the results validated and removed much faster, thereby minimizing the number of files on the RAID.

---

Hello Jim,

One benefit to the project of dropping the issuance of a 'initial 4th result' would be to decrease the number of files in the download directory by about 25%, assuming the number of issued work units is constant. This would help with the current problem of 'too many files' in the download directories, and give the project a little more room to grow on the current disk hardware.

Actually, I think it would have the opposite affect. Instead of decreasing the number of files, it would actually increase the number of files. The problem they found was that a large percentage of results were not being returned, therefore the WU's had to be reissued which added another 2 week delay. now we have those files on the RAID for 4 weeks (and sometimes more), whereas if they issue 4 initially, but only 3 get returned, they still get a quorum and the files can be deleted after the initial 2 weeks.

So, by speeding things up, they can get the results validated and removed much faster, thereby minimizing the number of files on the RAID.

Jim, of course you are right. Sending out four results helps to clear the file system faster. But: That is just another "profit" that SETI takes from the fourth results as stated in my initial posting. Just another reason that SETI must not "punish" the 4th-result-crunchers by ignoring their valid results.

Carsten