Where did all the LL assignments go?

petrw1 · 2006-11-27, 16:55

Sometime after Friday Nov 24 and before Monday Nov 27 almost 200 LL assignments in the 14M-17M ranges inclusive have "vanished" including some that were only assigned in the last few weeks. They are no longer in the assigned file (http://mersenne.org/primenet/status.txt) and did NOT move to the cleared file (http://mersenne.org/primenet/cleared.txt).

Here are the 45 "vanished" numbers in the 14M range: 14090623, 14108231, 14212379, 14257651, 14260271, 14271473, 14288077, 14325617, 14355337, 14360881, 14370313, 14395531, 14454361, 14488493, 14497309, 14534281, 14617357, 14625257, 14626897, 14651393, 14675987, 14679013, 14694191, 14717537, 14721281, 14737087, 14753279, 14759287, 14787139, 14798401, 14800559, 14801263, 14806249, 14814313, 14856763, 14857307, 14857721, 14893987, 14904047, 14908409
14919371, 14928521, 14989543, 14991997, 14997929

markr · 2006-11-28, 02:48

Nearly all of them (41 out of 45) had verified LL tests according to the GIMPS files of 9 November. The other four exponents are 14801263, 14806249, 14814313, 14928521 - I don't know what their story is.

Quite likely the whole group has been manually or programmatically removed from primenet in a cleanup of exponents that ought not be there. Many exponents expire & are reassigned repeatedly, and sometimes one of the previous assignees will check in much later. As long as it is in primenet's active ranges & not assigned to anybody else, primenet seems to accept it as active. I'm sure there are other ways this sort of thing happens, too.

Each range is eventually removed from priment, but cleaning up the trailing end is needed so the reports show the true progress. The forum regularly has questions like "I though every exponent below XX million was double-checked - why does the primenet status page show some below that?"

Like all cleaning, it's a never-ending job. The latest status file has a new crop of verified exponents: 14050301, 14223347, 14260271, 14264087, 14350387, ...

petrw1 · 2006-11-28, 04:14

Quote:

Originally Posted by markr

Nearly all of them (41 out of 45) had verified LL tests according to the GIMPS files of 9 November.

Like all cleaning, it's a never-ending job. The latest status file has a new crop of verified exponents: 14050301, 14223347, 14260271, 14264087, 14350387, ...

1. Then should not the Nov 15 synchronization have caught these?

2. In that case how can we be sure every exponent under 14M is, in fact done?

Uncwilly · 2006-11-28, 14:17

Quote:

Originally Posted by petrw1

2. In that case how can we be sure every exponent under 14M is, in fact done?

George keeps a human eye on the bottom end. Sometimes he does some manual things, like asking a known person to knock off a particular exponent that is on the bottom.

petrw1 · 2006-11-28, 16:16

Quote:

Originally Posted by Uncwilly

George keeps a human eye on the bottom end. Sometimes he does some manual things, like asking a known person to knock off a particular exponent that is on the bottom.

Makes sense....that would have been my guess.

How do I become a "known person"? It would seem like I'm a part of history: "The guy who completed the last exponent in the 14M range!"

garo · 2006-11-28, 21:27

George does the removal of exponents himself. I generate the list for him periodically based on the database files. All the 41 exponents are in my last list but I too don't know what the story with the other 4 is.

petrw1 · 2007-02-20, 18:12

Quote:

Originally Posted by garo

George does the removal of exponents himself. I generate the list for him periodically based on the database files. All the 41 exponents are in my last list but I too don't know what the story with the other 4 is.

Is this what happened again sometime in the last day or so? We went from over 30 LL tests in the 14M range to 1.

Prime95 · 2007-02-20, 20:21

Quote:

Originally Posted by petrw1

Is this what happened again sometime in the last day or so? We went from over 30 LL tests in the 14M range to 1.

Yes. Garo sent me another list of exponents that have been double-checked yet there was an open assignment. Those exponents have been purged from the server.

petrw1 · 2007-03-08, 21:01

Quote:

Originally Posted by Uncwilly

George keeps a human eye on the bottom end. Sometimes he does some manual things, like asking a known person to knock off a particular exponent that is on the bottom.

Is this why accounts SHANEAMY and DSWANSON get the majority of the assignments at "the bottom end"? For example of all the assignments in the 15M range these two have half of them and in the 16M range they have a third. If we throw in GARYFU and TEAM_PRIME_RIB the percentages go up to 61% and 44%. Further, excluding TPR, the other 3 have virtually NO assignments 17M and up.

AND...if I assume their prime (pun intended) role is cleaning up the "bottom end" and if someone (George?) manually makes it that way then it seems contradictory to that goal when I see odd re-assignment trickling in at the bottom end going to accounts other than these few. SHANEANY and DSWANSON rip through almost all remaining assignments in a short time and then these that trickle in later take several months more to finish off, some timing out and getting re-assigned again and again.

dswanson · 2007-03-16, 03:13

dswanson and shaneamy get no special treatment from George. We get our exponents the same way everyone else does: from the Primenet server. We're just particularly diligent at making sure we request exponents as soon as the server makes the expiring exponents available at 06:00 UTC.

We also make use of the various databases available to everyone to identify small exponents that need triplechecks but aren't assigned to anyone. We simply tell the server that we're working on them. As mentioned earlier in this thread, the server is perfectly happy to give us such exponents.

I think we've been very effective at moving the trailing edge of the doublechecks along at a reasonable pace. Because we wait for exponents to expire or otherwise become available, and because we don't poach exponents, there are always a few that get left behind because their owners continue to periodically check in, even though they are making little or no progress. The 10 or so exponents left in the 14M range fall into this category. In the past someone (I won't name names) poached the remaining few exponents in the 11M, 12M, and 13M ranges when the count got low. I'm expecting to see such behavior again soon.

petrw1 · 2007-03-18, 23:48

I hope my note didn't affend you ... I was responding to an earlier comment by uncwilly:

"George keeps a human eye on the bottom end. Sometimes he does some manual things, like asking a known person to knock off a particular exponent that is on the bottom."

I fully support his strategy.

Thanks for the honest response and thanks for the diligence.

It is sad that poaching happens ... and can't be stopped. Maybe I am slightly misinterpreting what poaching is ... is it when someone manually assigns themselves an active but slow-moving assignment and in the processs beats the rightful owner of the exponent to the credit? It seems to me that is something that could be monitored, so I am probably off the mark.

2006-11-27, 16:55	#1
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 2×3²×11×23 Posts	Where did all the LL assignments go? Sometime after Friday Nov 24 and before Monday Nov 27 almost 200 LL assignments in the 14M-17M ranges inclusive have "vanished" including some that were only assigned in the last few weeks. They are no longer in the assigned file (http://mersenne.org/primenet/status.txt) and did NOT move to the cleared file (http://mersenne.org/primenet/cleared.txt). Here are the 45 "vanished" numbers in the 14M range: 14090623, 14108231, 14212379, 14257651, 14260271, 14271473, 14288077, 14325617, 14355337, 14360881, 14370313, 14395531, 14454361, 14488493, 14497309, 14534281, 14617357, 14625257, 14626897, 14651393, 14675987, 14679013, 14694191, 14717537, 14721281, 14737087, 14753279, 14759287, 14787139, 14798401, 14800559, 14801263, 14806249, 14814313, 14856763, 14857307, 14857721, 14893987, 14904047, 14908409 14919371, 14928521, 14989543, 14991997, 14997929 Last fiddled with by petrw1 on 2006-11-27 at 17:01 Reason: Added paragraph 2 listing examples

2006-11-28, 02:48	#2
markr "Mark" Feb 2003 Sydney 3×191 Posts	Nearly all of them (41 out of 45) had verified LL tests according to the GIMPS files of 9 November. The other four exponents are 14801263, 14806249, 14814313, 14928521 - I don't know what their story is. Quite likely the whole group has been manually or programmatically removed from primenet in a cleanup of exponents that ought not be there. Many exponents expire & are reassigned repeatedly, and sometimes one of the previous assignees will check in much later. As long as it is in primenet's active ranges & not assigned to anybody else, primenet seems to accept it as active. I'm sure there are other ways this sort of thing happens, too. Each range is eventually removed from priment, but cleaning up the trailing end is needed so the reports show the true progress. The forum regularly has questions like "I though every exponent below XX million was double-checked - why does the primenet status page show some below that?" Like all cleaning, it's a never-ending job. The latest status file has a new crop of verified exponents: 14050301, 14223347, 14260271, 14264087, 14350387, ... Last fiddled with by markr on 2006-11-28 at 02:51

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
No assignments available?	dh1	GPU to 72	5	2016-10-10 03:36
Can't get new assignments	keiken	GPU to 72	2	2015-07-28 13:50
about assignments	savoie	Information & Answers	3	2015-04-30 10:07
assignments	mpeisl11	Information & Answers	4	2009-04-28 19:22
Old assignments	brunoparga	Operation Billion Digits	6	2006-12-28 20:34

2006-11-28, 21:27	#6
garo Aug 2002 Termonfeckin, IE ACB₁₆ Posts	George does the removal of exponents himself. I generate the list for him periodically based on the database files. All the 41 exponents are in my last list but I too don't know what the story with the other 4 is.

2007-03-16, 03:13	#10
dswanson Aug 2002 310₈ Posts	dswanson and shaneamy get no special treatment from George. We get our exponents the same way everyone else does: from the Primenet server. We're just particularly diligent at making sure we request exponents as soon as the server makes the expiring exponents available at 06:00 UTC. We also make use of the various databases available to everyone to identify small exponents that need triplechecks but aren't assigned to anyone. We simply tell the server that we're working on them. As mentioned earlier in this thread, the server is perfectly happy to give us such exponents. I think we've been very effective at moving the trailing edge of the doublechecks along at a reasonable pace. Because we wait for exponents to expire or otherwise become available, and because we don't poach exponents, there are always a few that get left behind because their owners continue to periodically check in, even though they are making little or no progress. The 10 or so exponents left in the 14M range fall into this category. In the past someone (I won't name names) poached the remaining few exponents in the 11M, 12M, and 13M ranges when the count got low. I'm expecting to see such behavior again soon.

2007-03-18, 23:48	#11
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 2×3²×11×23 Posts	I hope my note didn't affend you ... I was responding to an earlier comment by uncwilly: "George keeps a human eye on the bottom end. Sometimes he does some manual things, like asking a known person to knock off a particular exponent that is on the bottom." I fully support his strategy. Thanks for the honest response and thanks for the diligence. It is sad that poaching happens ... and can't be stopped. Maybe I am slightly misinterpreting what poaching is ... is it when someone manually assigns themselves an active but slow-moving assignment and in the processs beats the rightful owner of the exponent to the credit? It seems to me that is something that could be monitored, so I am probably off the mark.