Public PRPNet Servers

[rogue]

I am creating this thread to list the known public PRPNet servers hosted by various projects. These are the current servers that I am aware of. If you know of any corrections or additions for this list, please follow up in this thread. Eventually an admin will keep this post up to date.

The latest version of PRPNet is available from here.

Server: prpnet.primegrid.com

port: 12000 - Sophie Germain Prime Search: k*2^666666-1
port: 12001 - 121 Prime Search: 121*2^n+1, 121*2^n-1
port: 12002 - Factorial Prime Search: n!+1, n!-1
port: 12003 - Generalized Fermat Prime Search: b^65536+1
port: 12004 - Generalized Cullen/Woodall Prime Search: n*b^n+1, n*b^n-1
port: 12005 - Generalized Fermat Prime Search: b^32768+1
port: 12006 - 27121 Prime Search: 27*2^n+1, 27*2^n-1, 121*2^n+1
port: 12007 - Proth Prime Search Extended: k*b^n+1, 1200<k<10000, n>500000
port: 12008 - Primorial Prime Search: p#+1, p#-1

Server: pgllr.mine.nu

port: 9000 - Prime Sierpinski Prime Search
port: 10000 - Proth Prime Search Extended: k*b^n+1, 1200<k<10000, 130000<n<140000

Server: prime2u.com

port: 11000 - Proth Prime Search Extended: k*b^n+1, 1200<k<10000, 140000<n<150000

[TimSorbet]

Here are two more public PRPnet servers, for NPLB and CRUS respectively:
http://www.mersenneforum.org/showthread.php?t=12224
http://www.mersenneforum.org/showthread.php?t=12225

[rogue]

Originally Posted by Mini-Geek

Here are two more public PRPnet servers, for NPLB and CRUS respectively:
http://www.mersenneforum.org/showthread.php?t=12224
http://www.mersenneforum.org/showthread.php?t=12225

Using the format I had in the first post:

Server: www.noprimeleftbehind.net

port: 1300 - Sierpinski Base 22 Conjecture (CRUS): k*22^n+1
port: 9000 - Riesel Prime Search (NPLB): 400<k<600, 600000<n<1000000

I noticed that port 1300 is an older release of PRPNet. Since port 9000 is at 3.2.5, is there a plan to get port 1300 up to that release as well? It also seems that the user stats are messed up. I would recommend clearing those stats or rebuilding user stats from completed tests.

[mdettweiler]

Originally Posted by rogue

I noticed that port 1300 is an older release of PRPNet. Since port 9000 is at 3.2.5, is there a plan to get port 1300 up to that release as well? It also seems that the user stats are messed up. I would recommend clearing those stats or rebuilding user stats from completed tests.

Yes, I do plan to get that one upgraded to 3.2.6. Right now there's only one person crunching on 1300, and I'd rather not spring it on him without warning...I was planning to send him a PM to give him a heads-up, and then upgrade the server once I received such confirmation, but kinda dropped the ball on that. I'll see about doing that within the next day or two.

Regarding the stats, in what way are they messed up? They look fine to me.

[rogue]

Originally Posted by mdettweiler

Yes, I do plan to get that one upgraded to 3.2.6. Right now there's only one person crunching on 1300, and I'd rather not spring it on him without warning...I was planning to send him a PM to give him a heads-up, and then upgrade the server once I received such confirmation, but kinda dropped the ball on that. I'll see about doing that within the next day or two.

Regarding the stats, in what way are they messed up? They look fine to me.

It appears that Gary's user stats and Lennart's user stats are incorrect.

[mdettweiler]

Originally Posted by rogue

It appears that Gary's user stats and Lennart's user stats are incorrect.

Note that the server didn't always have large base 22 tests in it; initially, we ran comparatively smaller Sierpinski base 6 work on that port. Perhaps that caused an apparent disparity in the number of tests vs. points? (AFAIK, most of Lennart's points were from S6 work, and most of Gary's from S22, if that helps.) Or is there something else entirely that I'm not seeing here?

[enderak]

Server: pgllr.mine.nu
port: 9000 - Prime Sierpinski Prime Search

This is actually the Extended Sierpinski Problem, seaching composite k's, 78557 < k < 271129

[rogue]

Originally Posted by mdettweiler

Note that the server didn't always have large base 22 tests in it; initially, we ran comparatively smaller Sierpinski base 6 work on that port. Perhaps that caused an apparent disparity in the number of tests vs. points? (AFAIK, most of Lennart's points were from S6 work, and most of Gary's from S22, if that helps.) Or is there something else entirely that I'm not seeing here?

The numbers still look suspicious. If you recall, there was a bug with 2.x where user stats would et corrupted. After you upgrade to 3.2.6 you can fix this by deleting the records from the UserStats table then using this SQL statement to rebuild the table:

insert into userstats (userid, testsperformed, prpsfound, primesfound, gfndivisorsfound, totalscore)
(select distinct userid, count(*), 0, 0, 0, sum(decimallength * decimallength / 100000000)
from candidate,candidatetest
where candidate.candidatename = candidatetest.candidatename group by userid);

It isn't perfect, but it gets the job done quickly. There are some limitations with this statement as it is written.

1) Users won't get credit for tests that are no longer on the server. If you have archived the tests for base 22 (and only base 22), then you should be able to figure out the user stats from that data. I will help you if you need it.
2) It loses primes/prps found, but you can either update this statement manually to compute those values or you can update the table with additional SQL statements after running this one.
3) It does not exclude tests that have not been returned. You just need to exclude records from CandidateTest where the Residue is null.

[mdettweiler]

Originally Posted by rogue

The numbers still look suspicious. If you recall, there was a bug with 2.x where user stats would et corrupted. After you upgrade to 3.2.6 you can fix this by deleting the records from the UserStats table then using this SQL statement to rebuild the table:

insert into userstats (userid, testsperformed, prpsfound, primesfound, gfndivisorsfound, totalscore)
(select distinct userid, count(*), 0, 0, 0, sum(decimallength * decimallength / 100000000)
from candidate,candidatetest
where candidate.candidatename = candidatetest.candidatename group by userid);

It isn't perfect, but it gets the job done quickly. There are some limitations with this statement as it is written.

1) Users won't get credit for tests that are no longer on the server. If you have archived the tests for base 22 (and only base 22), then you should be able to figure out the user stats from that data. I will help you if you need it.
2) It loses primes/prps found, but you can either update this statement manually to compute those values or you can update the table with additional SQL statements after running this one.
3) It does not exclude tests that have not been returned. You just need to exclude records from CandidateTest where the Residue is null.

Okay, the server has now been upgraded. Regarding the SQL statement you provided, I only have two candidates in CandidateTest, both of which were pending at the time of upgrade and still are. Thus, that would essentially blank out the stats except for two lonely tests from Mathew Steine (in which case I may as well start afresh).

Hmm...I just thought of something. Was I by chance supposed to leave onlytest.removed in the server directory when I ran ./prpserver -u to have all those completed tests imported into CandidateTest? I moved that file out of the directory and into a subdirectory where I was putting the old 2.4.6 files for safekeeping prior to the upgrade.

Oh well. Even if I did keep onlytest.removed in there to be imported into the DB, that still wouldn't cover a lot of the tests--many of the old base 6 tests were done four or five times over when we were debugging the blank-residue problem way back when, and once that was all over the mess was just too much to sort through so I picked up all the server files, dumped them into a subdirectory, and started clean for base 22. So while the stats still count the base 6 tests, only the base 22 tests would have been on record in onlytest.removed.

I'm a little reluctant to blank out the user stats entirely considering that, since CRUS does not have a stats database a la NPLB, this is our only source of stats for this port. I don't suppose there's a way to figure out a ballpark figure of where Gary and Lennart's stats "should" be, so I could just fix them manually? (Or I suppose I could just delete Gary and Lennart's stats entirely and leave everyone else's apparently-OK figures be... )

[rogue]

Originally Posted by mdettweiler

Okay, the server has now been upgraded. Regarding the SQL statement you provided, I only have two candidates in CandidateTest, both of which were pending at the time of upgrade and still are. Thus, that would essentially blank out the stats except for two lonely tests from Mathew Steine (in which case I may as well start afresh).

Hmm...I just thought of something. Was I by chance supposed to leave onlytest.removed in the server directory when I ran ./prpserver -u to have all those completed tests imported into CandidateTest? I moved that file out of the directory and into a subdirectory where I was putting the old 2.4.6 files for safekeeping prior to the upgrade.

Oh well. Even if I did keep onlytest.removed in there to be imported into the DB, that still wouldn't cover a lot of the tests--many of the old base 6 tests were done four or five times over when we were debugging the blank-residue problem way back when, and once that was all over the mess was just too much to sort through so I picked up all the server files, dumped them into a subdirectory, and started clean for base 22. So while the stats still count the base 6 tests, only the base 22 tests would have been on record in onlytest.removed.

I'm a little reluctant to blank out the user stats entirely considering that, since CRUS does not have a stats database a la NPLB, this is our only source of stats for this port. I don't suppose there's a way to figure out a ballpark figure of where Gary and Lennart's stats "should" be, so I could just fix them manually? (Or I suppose I could just delete Gary and Lennart's stats entirely and leave everyone else's apparently-OK figures be... )

You would have to compute stats from onlytest.removed. Those are not imported into the database. You could import them by renaming it to prpserver.candidates (they have the same format) then using the -u option when starting the server. I haven't tried that myself, but it should work (in theory).

Since it is a "base 22" server and since the base 6 stats were amuck, it should be fine to only include base 22 in the stats. You should ask Gary and Lennart if they care about their stats on that particular server. Computing an estimate for them might not be too difficult since you know how many tests they completed and probably have an average decimal length for those tests.

[mdettweiler]

Originally Posted by rogue

You would have to compute stats from onlytest.removed. Those are not imported into the database. You could import them by renaming it to prpserver.candidates (they have the same format) then using the -u option when starting the server. I haven't tried that myself, but it should work (in theory).

Since it is a "base 22" server and since the base 6 stats were amuck, it should be fine to only include base 22 in the stats. You should ask Gary and Lennart if they care about their stats on that particular server. Computing an estimate for them might not be too difficult since you know how many tests they completed and probably have an average decimal length for those tests.

Ah, I see. Okay, then, in that case what I'll do is use the # of tests figures to come up with a rough estimate of how many points each should have (assuming all base 22 for Gary and all base 6 for Lennart, which is about right). Their scores are both far enough ahead of everyone else that a few points' difference shouldn't be too big a deal.