## Other

drummers-lowrise

Message boards : Prime Sierpinski Problem : Completed, waiting for validation

 Subscribe SortOldest firstNewest firstHighest rated posts first
Author Message
ptrnyc

Joined: 8 Jul 17
Posts: 3
ID: 905395
Credit: 208,783
RAC: 0

Message 109046 - Posted: 23 Jul 2017 | 20:15:39 UTC

How long do WU typically stay in that state ? I have several pending, for example 533298218 which has been stuck there for a week now....

Crun-chi
Volunteer tester

Joined: 25 Nov 09
Posts: 2926
ID: 50683
Credit: 56,427,743
RAC: 10,773

Message 109048 - Posted: 23 Jul 2017 | 20:50:19 UTC - in response to Message 109046.

How long do WU typically stay in that state ? I have several pending, for example 533298218 which has been stuck there for a week now....

Sometimes it can took while until finished, but you dont need to worry, just finished and enjoy. Credits will come :)

____________
314187728^131072+1 GENERALIZED FERMAT :)
93*10^1029523-1 REPDIGIT PRIME
31*332^367560+1 CRUS PRIME
Proud member of team Aggie The Pew. Go Aggie!

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13260
ID: 53948
Credit: 223,143,742
RAC: 32,556

Message 109049 - Posted: 23 Jul 2017 | 21:07:32 UTC - in response to Message 109046.

How long do WU typically stay in that state ? I have several pending, for example 533298218 which has been stuck there for a week now....

All tasks at PrimeGrid must be verified by at least two computers before they can be validated. Yours is one computer, and a second computer must also return a matching result.

You're waiting for the other computer (aka your "wingman") to return the result. Some computers are fast. Some are slow. Some people run their computers 24 hours a day, some run them only an hour or two each day. Some computers are on 7 days a week, while others are on only during the work week and yet others only on the weekend. PSP tasks have a deadline of 21 days, and in the right circumstances that deadline will be extended to as much as 84 days.

That, of course, assumes your wingman actually returns a valid result. If he just turns off his computer, after 21 days the server will send out a replacement task to a third computer, and then we wait for that computer to return the task. If his computer has a calculation error, a third task needs to be sent out.

If you get unlucky with wingmen, it can be a while before the task finally gets validated.

So waiting a week is hardly a long time for PSP, all things considered. Currently, the longest running active PSP workunit has been in progress for a little over three months.

____________
My lucky number is 75898524288+1

ptrnyc

Joined: 8 Jul 17
Posts: 3
ID: 905395
Credit: 208,783
RAC: 0

Message 109051 - Posted: 23 Jul 2017 | 22:09:24 UTC - in response to Message 109049.

Damn. Right now I'm running PSP 24/7 and crunching 2 WU every 3 days - way faster than they get validated... maybe I'll slow it down a bit...

Rafael
Volunteer tester

Joined: 22 Oct 14
Posts: 859
ID: 370496
Credit: 312,846,449
RAC: 49,358

Message 109052 - Posted: 23 Jul 2017 | 22:19:39 UTC - in response to Message 109051.

Damn. Right now I'm running PSP 24/7 and crunching 2 WU every 3 days - way faster than they get validated... maybe I'll slow it down a bit...

Crunching faster than they can validate is actually a good thing: if you were to get lucky enough to receive a prime number, you'd be credited as the prime finder, and not just the double checker. Don't worry, everything will eventually get checked and rewards applied. Just wait and it'll happen.

ptrnyc

Joined: 8 Jul 17
Posts: 3
ID: 905395
Credit: 208,783
RAC: 0

Message 109053 - Posted: 23 Jul 2017 | 22:21:21 UTC - in response to Message 109052.

Thanks.

dlawson

Joined: 7 Jun 17
Posts: 25
ID: 895562
Credit: 67,807,749
RAC: 4,936

Message 110537 - Posted: 29 Sep 2017 | 15:15:06 UTC - in response to Message 109053.

My PSP work unit was awaiting validation, the other user aborted it and I guess the system didn't send it out again because it was of the same batch where the prime was recently found.

I am all for focusing the energy in the right places but wonder what will happen to this case, which could be another prime (optimistic I know!!)

http://www.primegrid.com/workunit.php?wuid=538667682

Thanks for your collective wisdom!

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13260
ID: 53948
Credit: 223,143,742
RAC: 32,556

Message 110538 - Posted: 29 Sep 2017 | 15:30:05 UTC - in response to Message 110537.

My PSP work unit was awaiting validation, the other user aborted it and I guess the system didn't send it out again because it was of the same batch where the prime was recently found.

I am all for focusing the energy in the right places but wonder what will happen to this case, which could be another prime (optimistic I know!!)

http://www.primegrid.com/workunit.php?wuid=538667682

Thanks for your collective wisdom!

Your task was actually a different K than the PSP prime that was found, so there's nothing at all unusual about this workunit. However, even if it was the same K, this workunit would still be run to completion. As you say, it could be prime and there's no way of knowing without checking. We also don't want your work to go to waste. We won't normally stop a workunit in the middle unless there's something wrong with it.

At any given instant in time, the system queues up about 400 PSP tasks that can be sent out. It does this for all subprojects, although the number of tasks in the queue varies from 200 to 4000 depending on the subproject. This way there's always work available for your computers.

While the tasks are sitting in that queue waiting to be sent out, they show up as "unsent". The particular task you're looking at will eventually be sent out to a computer for processing.
____________
My lucky number is 75898524288+1

dlawson

Joined: 7 Jun 17
Posts: 25
ID: 895562
Credit: 67,807,749
RAC: 4,936

Message 110543 - Posted: 29 Sep 2017 | 19:04:03 UTC - in response to Message 110538.

Thank you for the helpful explanation.
I tried to check the k value but couldn't figure it out.
Is there a way to do it for normal users like me?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13260
ID: 53948
Credit: 223,143,742
RAC: 32,556

Message 110550 - Posted: 29 Sep 2017 | 19:58:19 UTC - in response to Message 110543.

Thank you for the helpful explanation.
I tried to check the k value but couldn't figure it out.
Is there a way to do it for normal users like me?

While the task is running, you can look at the input file that's sent as part of the task. (I'll leave it as an exercise to the reader to figure out the details.)
____________
My lucky number is 75898524288+1

composite
Volunteer tester

Joined: 16 Feb 10
Posts: 752
ID: 55391
Credit: 671,437,205
RAC: 370,795

Message 110561 - Posted: 30 Sep 2017 | 16:05:36 UTC - in response to Message 110550.

Which leads one to wonder if conspirators can raise their total prime scores by inducing the system to falsely return any given workunit they have in common as prime (particularly the uninteresting ones not reportable to T5K), by entering known a prime into the input file after the workunit has been received, but before it has been processed.

Is the database checking that the test reported in the output file matches the inputs sent with the workunit?

Where is the proof of work? Is the output cryptographically hashed with the input to prevent tampering with the files?

I just want to know that the research work we are submitting in good faith is not being tainted by hacking.

Dirk Sellsted

Joined: 15 Feb 17
Posts: 151
ID: 492091
Credit: 139,650,217
RAC: 24,860

Message 110564 - Posted: 30 Sep 2017 | 17:40:11 UTC - in response to Message 110561.

files are sent to multiple hosts, I can't be my own double checker. Also JimB i believe will run suspicious results himself. I think there's a sgs prime residue .0001 or something which is almost always a false positive.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13260
ID: 53948
Credit: 223,143,742
RAC: 32,556

Message 110568 - Posted: 30 Sep 2017 | 18:18:59 UTC - in response to Message 110561.

Which leads one to wonder if conspirators can raise their total prime scores by inducing the system to falsely return any given workunit they have in common as prime (particularly the uninteresting ones not reportable to T5K), by entering known a prime into the input file after the workunit has been received, but before it has been processed.

Is the database checking that the test reported in the output file matches the inputs sent with the workunit?

Where is the proof of work? Is the output cryptographically hashed with the input to prevent tampering with the files?

I just want to know that the research work we are submitting in good faith is not being tainted by hacking.

tl;dr: One could try, but in practice it wouldn't work.

Large LLR tasks ultimately get checked somewhere, either at PrimeGrid or T5K.

All GFN tasks, large or small, must be checked by us anyway because you're only doing a PRP test and we still need to run the primality test.

That leaves just SGS, and if we were to get flooded with false SGS primes, it would be immediately obvious to Jim or myself. To do what you're talking about, you would need to submit hundreds or thousands of false primes. Our alert level goes up when we see 3 or 4 false primes. We're looking for malfunctions, not cheaters, but anyone trying to cheat would be blatantly obvious to us.
____________
My lucky number is 75898524288+1

composite
Volunteer tester

Joined: 16 Feb 10
Posts: 752
ID: 55391
Credit: 671,437,205
RAC: 370,795

Message 110569 - Posted: 30 Sep 2017 | 18:27:09 UTC - in response to Message 110564.

files are sent to multiple hosts, I can't be my own double checker.
It requires a team of conspirators to try capturing some fresh workunits. This is not an easy thing to do in a high-volume application.
Also JimB i believe will run suspicious results himself.
That's the point of staying under the radar, avoiding results that are going to attract attention and scrutiny.
I think there's a sgs prime residue .0001 or something which is almost always a false positive.
Possibly, but I'm talking about editing the input file, putting a genuine known prime in there to test. It won't take any less work to test than a genuine prime, and it will return a real positive. The theory is that the result is associated with a workunit for a number whose prime status is unknown, and it will remain silently unknown if the system accepts the fake prime report for that workunit and considers it as the result for the original number that should be tested. The only guaranteed protection is to retest (offline) every prime that is reported and verified through BOINC.

composite
Volunteer tester

Joined: 16 Feb 10
Posts: 752
ID: 55391
Credit: 671,437,205
RAC: 370,795

Message 110571 - Posted: 30 Sep 2017 | 18:31:42 UTC - in response to Message 110568.
Last modified: 30 Sep 2017 | 19:05:49 UTC

To do what you're talking about, you would need to submit hundreds or thousands of false primes. Our alert level goes up when we see 3 or 4 false primes.
I see, it's very unlikely that people would take the trouble to slip in a fake result every now and then.

EDIT: Anyone want to try a proof-of-concept with SGS?
EDIT EDIT: It becomes a lot harder to get 3 conspirators when replication is 3. But that wastes a lot of processing power when the goal is to catch errors. Post-validation of positive results is better to catch cheaters.

Dirk Sellsted

Joined: 15 Feb 17
Posts: 151
ID: 492091
Credit: 139,650,217
RAC: 24,860

Message 110574 - Posted: 30 Sep 2017 | 19:39:51 UTC - in response to Message 110571.

If you were to replace the actual input file with a known SGS prime then it would be caught when they added it to their list of known primes?

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,118,817
RAC: 0

Message 110577 - Posted: 30 Sep 2017 | 21:08:20 UTC
Last modified: 30 Sep 2017 | 21:09:21 UTC

The validator is searching for the exact candidate we sent out. If the upload is for a different candidate it's the same as if the upload was missing. It'll be marked as a validate error. For sieving everything about every factor is checked: p range, k value, n range, b value, c value, whether it really divides the candidate.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13260
ID: 53948
Credit: 223,143,742
RAC: 32,556

Message 110578 - Posted: 30 Sep 2017 | 21:10:02 UTC
Last modified: 30 Sep 2017 | 21:13:14 UTC

There are plenty of safeguards in place to catch malfunctioning systems -- which are probably a lot more devious than someone trying to cheat! You wouldn't believe the crazy stuff some computers send back.

Our validator is very thorough and very sophisticated.

With the exception of SGS primes that aren't reported to T5K, all primes are verified by either PrimeGrid's computers or T5K's computers. It's therefore impossible to "fake" a prime because we check them all ourselves. We haven't decided yet whether it's necessary to verify the "too-small" SGS primes, but if we do, it would be a trivial exercise to retroactively test the primes already in the database. It's not something that overly concerns me -- and I'm generally pretty paranoid about cheating.
____________
My lucky number is 75898524288+1

Renix

Joined: 26 Aug 16
Posts: 283
ID: 455383
Credit: 1,720,171,040
RAC: 3,980,973

Message 110579 - Posted: 30 Sep 2017 | 22:05:44 UTC

just thoughts on cheater... I wouldn't think really that there would be many people trying to cheat here. After all, if they are after points, there's projects that do more points. As for cheating to find primes I think that would be quite hard to do. Too many log files, servers sending out work, work units tracked, results turned back in being "prime" go through further checks. An awful lot of work. This project has a heck of a "paper trail" so to speak.

Dirk Sellsted

Joined: 15 Feb 17
Posts: 151
ID: 492091
Credit: 139,650,217
RAC: 24,860

Message 110581 - Posted: 1 Oct 2017 | 1:17:39 UTC

I'd encourage anyone to try to cheat and find a way through the system, and to let michael or jim know how they did it. However I doubt anyone would succeed.

Rafael
Volunteer tester

Joined: 22 Oct 14
Posts: 859
ID: 370496
Credit: 312,846,449
RAC: 49,358

Message 110582 - Posted: 1 Oct 2017 | 1:34:23 UTC - in response to Message 110581.

I'd encourage anyone to try to cheat and find a way through the system, and to let michael or jim know how they did it. However I doubt anyone would succeed.

That's easy, just develop new math that allows you to predict primes / residues so that you can easily pinpoint primes by hand instead of going for brute force methods like we do right now.

Kappa.

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 13260
ID: 53948
Credit: 223,143,742
RAC: 32,556

Message 110584 - Posted: 1 Oct 2017 | 5:41:49 UTC - in response to Message 110581.

I'd encourage anyone to try to cheat and find a way through the system, and to let michael or jim know how they did it. However I doubt anyone would succeed.

That's a really good way to get yourself permanently banned from PrimeGrid, along with forfeiting all credit. Don't even joke about this.
____________
My lucky number is 75898524288+1

Dirk Sellsted

Joined: 15 Feb 17
Posts: 151
ID: 492091
Credit: 139,650,217
RAC: 24,860

Message 110585 - Posted: 1 Oct 2017 | 5:51:03 UTC - in response to Message 110584.

Wasn't a joke but a sincere form of black box testing. However it was written in a softer mood that doesn't convey over text I'm sure. Feel free to delete that and subsequent messages regarding the subject.

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 916
ID: 107307
Credit: 974,118,817
RAC: 0

Message 110591 - Posted: 1 Oct 2017 | 11:01:04 UTC

Just to prove a point, I retested every prime found in the SGS search that was not rechecked by the Top 5000 Primes site:

3491091900915*2^1290000-1 is prime! (388342 decimal digits) Time : 370.384 sec.
3661983716277*2^1290000-1 is prime! (388342 decimal digits) Time : 370.356 sec.
3662671972137*2^1290000-1 is prime! (388342 decimal digits) Time : 367.467 sec.
3653167449585*2^1290000-1 is prime! (388342 decimal digits) Time : 369.563 sec.
3664698954765*2^1290000-1 is prime! (388342 decimal digits) Time : 370.819 sec.
3665216176377*2^1290000-1 is prime! (388342 decimal digits) Time : 371.723 sec.
3664348119747*2^1290000-1 is prime! (388342 decimal digits) Time : 370.981 sec.
3666180071517*2^1290000-1 is prime! (388342 decimal digits) Time : 370.391 sec.
3672385818237*2^1290000-1 is prime! (388342 decimal digits) Time : 367.145 sec.
3672837975165*2^1290000-1 is prime! (388342 decimal digits) Time : 371.933 sec.
3672793921977*2^1290000-1 is prime! (388342 decimal digits) Time : 361.624 sec.
3671950858035*2^1290000-1 is prime! (388342 decimal digits) Time : 351.408 sec.
3671874087807*2^1290000-1 is prime! (388342 decimal digits) Time : 351.115 sec.
3674516188887*2^1290000-1 is prime! (388342 decimal digits) Time : 350.927 sec.
3667352439987*2^1290000-1 is prime! (388342 decimal digits) Time : 350.905 sec.
3677218438005*2^1290000-1 is prime! (388342 decimal digits) Time : 351.703 sec.
3676174985145*2^1290000-1 is prime! (388342 decimal digits) Time : 350.716 sec.
3679206502125*2^1290000-1 is prime! (388342 decimal digits) Time : 350.569 sec.
3681118197447*2^1290000-1 is prime! (388342 decimal digits) Time : 351.685 sec.
3679602409875*2^1290000-1 is prime! (388342 decimal digits) Time : 351.528 sec.
3681233023275*2^1290000-1 is prime! (388342 decimal digits) Time : 350.397 sec.
3678629578785*2^1290000-1 is prime! (388342 decimal digits) Time : 351.272 sec.
3683871972357*2^1290000-1 is prime! (388342 decimal digits) Time : 351.058 sec.
3684297411765*2^1290000-1 is prime! (388342 decimal digits) Time : 350.372 sec.
3677587837107*2^1290000-1 is prime! (388342 decimal digits) Time : 350.948 sec.
3684885397455*2^1290000-1 is prime! (388342 decimal digits) Time : 351.321 sec.
3685163118927*2^1290000-1 is prime! (388342 decimal digits) Time : 351.772 sec.
3686585508177*2^1290000-1 is prime! (388342 decimal digits) Time : 351.081 sec.
3673491804495*2^1290000-1 is prime! (388342 decimal digits) Time : 351.183 sec.
3681723548067*2^1290000-1 is prime! (388342 decimal digits) Time : 351.374 sec.
3693825100077*2^1290000-1 is prime! (388342 decimal digits) Time : 351.746 sec.
3688019625645*2^1290000-1 is prime! (388342 decimal digits) Time : 350.689 sec.
3695646799815*2^1290000-1 is prime! (388342 decimal digits) Time : 351.638 sec.
3697934149005*2^1290000-1 is prime! (388342 decimal digits) Time : 351.152 sec.
3699486130767*2^1290000-1 is prime! (388342 decimal digits) Time : 350.076 sec.
3700672944267*2^1290000-1 is prime! (388342 decimal digits) Time : 351.313 sec.
3700402879665*2^1290000-1 is prime! (388342 decimal digits) Time : 351.143 sec.
3701677179585*2^1290000-1 is prime! (388342 decimal digits) Time : 350.522 sec.
3704349928587*2^1290000-1 is prime! (388342 decimal digits) Time : 350.508 sec.
3705035506065*2^1290000-1 is prime! (388342 decimal digits) Time : 351.845 sec.
3705816430155*2^1290000-1 is prime! (388342 decimal digits) Time : 351.402 sec.
3706586779815*2^1290000-1 is prime! (388342 decimal digits) Time : 355.257 sec.
3704839591605*2^1290000-1 is prime! (388342 decimal digits) Time : 355.610 sec.
3706451204967*2^1290000-1 is prime! (388342 decimal digits) Time : 354.009 sec.
3700172480067*2^1290000-1 is prime! (388342 decimal digits) Time : 355.159 sec.
3708773808777*2^1290000-1 is prime! (388342 decimal digits) Time : 355.754 sec.
3689340559557*2^1290000-1 is prime! (388342 decimal digits) Time : 353.740 sec.
3712091764047*2^1290000-1 is prime! (388342 decimal digits) Time : 355.144 sec.
3712800178425*2^1290000-1 is prime! (388342 decimal digits) Time : 355.783 sec.
3712782951717*2^1290000-1 is prime! (388342 decimal digits) Time : 355.417 sec.
3713907919935*2^1290000-1 is prime! (388342 decimal digits) Time : 353.553 sec.
3719462989305*2^1290000-1 is prime! (388342 decimal digits) Time : 354.670 sec.
3723164887647*2^1290000-1 is prime! (388342 decimal digits) Time : 355.579 sec.
3723571452645*2^1290000-1 is prime! (388342 decimal digits) Time : 353.781 sec.
3722496953565*2^1290000-1 is prime! (388342 decimal digits) Time : 355.678 sec.
3722645982195*2^1290000-1 is prime! (388342 decimal digits) Time : 356.135 sec.
3720299748165*2^1290000-1 is prime! (388342 decimal digits) Time : 354.755 sec.
3720255807765*2^1290000-1 is prime! (388342 decimal digits) Time : 355.576 sec.
3721745642685*2^1290000-1 is prime! (388342 decimal digits) Time : 355.885 sec.
3720800327307*2^1290000-1 is prime! (388342 decimal digits) Time : 356.992 sec.
3720815370627*2^1290000-1 is prime! (388342 decimal digits) Time : 374.902 sec.
3725634193125*2^1290000-1 is prime! (388342 decimal digits) Time : 377.613 sec.
3727652988237*2^1290000-1 is prime! (388342 decimal digits) Time : 383.206 sec.
3722796060027*2^1290000-1 is prime! (388342 decimal digits) Time : 376.648 sec.
3729146630055*2^1290000-1 is prime! (388342 decimal digits) Time : 377.809 sec.
3728408685045*2^1290000-1 is prime! (388342 decimal digits) Time : 371.055 sec.
3732425988135*2^1290000-1 is prime! (388342 decimal digits) Time : 371.057 sec.
3732066748137*2^1290000-1 is prime! (388342 decimal digits) Time : 370.928 sec.
3734173086327*2^1290000-1 is prime! (388342 decimal digits) Time : 369.307 sec.
3734856081225*2^1290000-1 is prime! (388342 decimal digits) Time : 364.074 sec.
3735207275037*2^1290000-1 is prime! (388342 decimal digits) Time : 367.088 sec.

Please note that they're all really prime. Exactly as expected.

composite
Volunteer tester

Joined: 16 Feb 10
Posts: 752
ID: 55391
Credit: 671,437,205
RAC: 370,795

Message 110600 - Posted: 1 Oct 2017 | 15:59:57 UTC - in response to Message 110582.
Last modified: 1 Oct 2017 | 16:00:59 UTC

just develop new math that allows you to predict primes / residues so that you can easily pinpoint primes by hand instead of going for brute force methods like we do right now.

Kappa.

That's not cheating, that would be a genuine revolutionary scientific breakthrough.

Renix

Joined: 26 Aug 16
Posts: 283
ID: 455383
Credit: 1,720,171,040
RAC: 3,980,973

Message 110601 - Posted: 1 Oct 2017 | 20:08:55 UTC - in response to Message 110591.

Just to prove a point, I retested every prime found in the SGS search that was not rechecked by the Top 5000 Primes site:

Please note that they're all really prime. Exactly as expected.

That's cool, just more verification about how solid the search program is. :-)

Message boards : Prime Sierpinski Problem : Completed, waiting for validation