The Primegrid monopoly

The Carnivore · 2010-09-15, 22:53

I know I'm potentially stirring up a hornet's nest with this post, but here it goes...

In all my years of watching and participating in the distributed computing world, I've never seen anything like Primegrid's takeover of other projects. Some of those takeovers were justified, but a few of them had nasty consequences that must be said, even if it's probably too late to do anything about it. Let's see some examples:

Three years ago, the PCP (Prime Cullen Prime) project was taken over:
http://www.mersenneforum.org/forumdisplay.php?f=79
That project searched for prime cullen primes, but instead of removing the prime cullen candidates from their Cullen search, Primegrid went ahead and tested all cullen candidates anyway, regardless of whether they were prime or not. Unfortunately, the PCP project was not the only casualty in Primegrid's ruthless march to domination.

A year later, Primegrid apparently decided that it wasn't enough to destroy one project, it had to get rid of another. On September 2008, the 3*2^n-1 project fell victim to Primegrid:
http://www.mersenneforum.org/forumdisplay.php?f=14

More conquests followed. The proth search project was taken over that same year:
http://www.prothsearch.net/index.html
Yves Gallot's Generalized Fermat Prime Search was absorbed by Primegrid soon after:
http://yves.gallot.pagesperso-orange.fr/primes/index.html
as was the 12121 search:
http://primes.utm.edu/bios/page.php?id=640

Today, Primegrid has become more cunning and is adopting a more subtle strategy. The 17 or bust project is getting swallowed up this time. It is already hard, if not impossible, to reserve a 17 or bust sieve range outside of Primegrid:
http://www.free-dc.org/forum/showthread.php?3501-Sieve-Information-getting-started
and the primality testing part of the project is slowly being transferred over: http://www.free-dc.org/forum/showthread.php?25274-Now-a-predominantly-PrimeGrid-project I don't know when this will happen, but it's only a matter of time before 17 or bust also succumbs to Primegrid.

Now I'm not saying that all of Primegrid's takeovers were wrong. Like I said before, some of those takeovers were justified, like when Primegrid took over Riesel Sieve since the admins of that project disappeared. But they should respect the rights and interests of other active projects, no matter how large or small they are. So don't let them take over another active effort again, even if you're not directly involved in it. If you're mainly an NPLB cruncher, join the RPS folks and speak up if Primegrid starts to test k<300 for riesel numbers. If you're mainly an RPS cruncher, join the NPLB folks and speak up if Primegrid starts to test riesel numbers for k=301-3200. And refuse to crunch for Primegrid if they ever do something like that. It may be too late to fix past wrongs, but we still can prevent future injustices.

rogue · 2010-09-15, 23:13

I am writing this to refute much of what you say. I maintain most of the projects on ProthSearch. I worked with PrimeGrid to help them get started. I also helped them get started with the Cullen and Woodall searches, which also originated on ProthSearch and had almost no remaining participants. The PCP project happens to be a subset of the Cullen Prime search. If you read this, then you will see that Rytis extends his regrets over it.

I also helped them restart the GFN search. Yves Gallot has been MIA for years and that project has not moved forward since 2004 (?). There have been private efforts to extend the GFN search, but they remained private and closed to others. David Underbakke (writer of genefx64) and I worked together so that genefx64 could be used by PrimeGrid via PRPNet. David was probably the only person actively searching for GFNs.

I can understand some people not being happy about PrimeGrid, but trust me, PrimeGrid has done its utmost to not step on the toes of other projects. The offer a stable and robust platform and server and it is easy to see why projects would want to get on their bandwagon.

I personally believe that PrimeGrid is one of the best distributed computing projects out there because it has greatly increased visibility to the searches they are undertaking. Most of these other projects were languishing and would have died completely if it weren't for PrimeGrid.

Note that PrimeGrid has avoided NPLB (No Prime Left Behind). It has also stayed away from TRP (The Riesel Problem) for many years due to the owners insisting that they would bring it back.

mdettweiler · 2010-09-15, 23:46

Quote:

Originally Posted by rogue

I am writing this to refute much of what you say. I maintain most of the projects on ProthSearch. I worked with PrimeGrid to help them get started. I also helped them get started with the Cullen and Woodall searches, which also originated on ProthSearch and had almost no remaining participants. The PCP project happens to be a subset of the Cullen Prime search. If you read this, then you will see that Rytis extends his regrets over it.

I also helped them restart the GFN search. Yves Gallot has been MIA for years and that project has not moved forward since 2004 (?). There have been private efforts to extend the GFN search, but they remained private and closed to others. David Underbakke (writer of genefx64) and I worked together so that genefx64 could be used by PrimeGrid via PRPNet. David was probably the only person actively searching for GFNs.

I can understand some people not being happy about PrimeGrid, but trust me, PrimeGrid has done its utmost to not step on the toes of other projects. The offer a stable and robust platform and server and it is easy to see why projects would want to get on their bandwagon.

I personally believe that PrimeGrid is one of the best distributed computing projects out there because it has greatly increased visibility to the searches they are undertaking. Most of these other projects were languishing and would have died completely if it weren't for PrimeGrid.

Note that PrimeGrid has avoided NPLB (No Prime Left Behind). It has also stayed away from TRP (The Riesel Problem) for many years due to the owners insisting that they would bring it back.

I agree that any fears that PrimeGrid will "take over" NPLB or RPS are quite unfounded. Gary and I (the NPLB admins) have a working rapport with the PrimeGrid admins and neither project is interested in taking over the other. All of PrimeGrid's efforts were either a) effectively dead when they took them over or b) merged by the explicit and enthusiastic consent of the project admins. NPLB and RPS are both active projects that are a far cry from "dead" and thus nobody at PrimeGrid would think of taking them over. Both projects have internally discussed the possibility of going BOINC in the past (with the option of doing so through PrimeGrid having been floated around), but for various reasons decided not to at this time and PrimeGrid has respected that.

I, in company with various other NPLB members, crunch for PrimeGrid at times, as do various PrimeGrid members. Lennart, one of the admins at PrimeGrid, has lent a huge hand at NPLB with his massive computing resources during our rallies. PrimeGrid even allows us to announce our rallies in their forum, and we would be equally open to the reverse if it were proposed. Clearly the projects are neither at odds with or plotting takeovers of each other.

Max

henryzz · 2010-09-16, 19:13

The reason a lot of projects lose activity and/or change to primegrid is because the tests for most projects get larger and larger and take longer and longer. The only way to make any significant progess after a while is to turn to BOINC. It would surprise me if at least one of NPLB, RPS, or CRUS didn't end up asking Primegrid or BOINC for help in the next few years. I expect that the projects would all remain working here on mersenneforum as well because of the nature of the projects.

kar_bon · 2010-09-16, 19:19

Quote:

Originally Posted by henryzz

The reason a lot of projects lose activity and/or change to primegrid is because the tests for most projects get larger and larger and take longer and longer. The only way to make any significant progess after a while is to turn to BOINC.

Why? There's no need to use BOINC only because the tests last longer.
PRPNet and LLRnet can handle longer runtimes. It's a setting in the server configuration only, to give the client the time needed to complete a test.
For example, all servers at NPLB are set to two days, which give the client 2 days to complete a test before the server will hand out that test to another client.

mdettweiler · 2010-09-16, 19:34

Quote:

Originally Posted by kar_bon

Why? There's no need to use BOINC only because the tests last longer.
PRPNet and LLRnet can handle longer runtimes. It's a setting in the server configuration only, to give the client the time needed to complete a test.
For example, all servers at NPLB are set to two days, which give the client 2 days to complete a test before the server will hand out that test to another client.

I think what henryzz is saying is that as projects' test sizes get bigger, fewer people want to crunch them, and the project regulars can't move the ranges ahead at a decently fast rate; hence PrimeGrid opens up an influx of new participants to revitalize the flow of work.

As far as NPLB and CRUS go, though, the nature of the projects is such that I don't forsee either running into such a problem. Projects like PSP, SoB, and 321 have one or a small handful of k's that they're taking to ever-increasing n-levels; NPLB and CRUS, though, are of much broader scope in terms of k or b, respectively. Both are pretty well spread out across the spectrum of test sizes, so they're never at a loss for "easy" work that's attainable without the huge resources of BOINC.

xilman · 2010-09-16, 19:39

Quote:

Originally Posted by mdettweiler

I think what henryzz is saying is that as projects' test sizes get bigger, fewer people want to crunch them, and the project regulars can't move the ranges ahead at a decently fast rate; hence PrimeGrid opens up an influx of new participants to revitalize the flow of work.

[rds]You are referring to the instant gratification crowd?[/rds]

Sorry, but I couldn't resist the opportunity of getting my retaliation in first.

In reality, I've nothing against the IGC --- quite the reverse. They do a lot of tedious grunt work and let me concentrate my resources on the harder and (IMAO) more interesting problems. For providing that benefit they deserve some praise.

Paul

gd_barnes · 2010-09-16, 22:12

Contrary to what most people might expect, I am in favor of possibly "BOINCing" a couple of the bases on CRUS in the distant future; that is "asking" PrimeGrid to share the effort with us. I can foresee a day when one of our bases reaches n=4M or 5M and hasn't had any activity in 6 months or more that it might make sense.

As a good example, if we are down to one final k remaining on Riesel base 6 (currently 2 k's are remaining, 1 at n=1M and 1 at n=820K) and we've searched it to n=4M with little interest shown for an extented period, that would be the time to ask PrimeGrid for help. At that point, the effort becomes virtually the equivalent of the 3*2^n-1 effort where one k is being searched to an extremely deep depth.

In response to the "instant gratification crowd", I certainly am one. The case that I make for doing it is that it is somewhat thankless work that few people want to do. There is no glory and even low-to-mid top-5000 work usually drops off within a year. The most thankless of all is non-top-5000 work but I feel it is worthwhile to have a full and accurate listing of primes.

Oddball · 2011-03-06, 19:37

Just wanted to point out that Primegrid now has more primes on the top 5000 list (2207) than projects #2-20 combined (2085):

http://primes.utm.edu/bios/top20.php...&by=PrimesRank

That may change over the next few months when their n=666666 primes begin dropping off the list.

ellipse · 2011-04-24, 20:55

Right now, Primegrid has 2391 primes on the top 5000 list, and the n=666666 primes have just started dropping off. The top 3 projects (Primegrid, RPS, and NPLB) currently have 3750 primes on that list. This means that exactly 75% of the primes on the list belongs to one of those three projects.

2010-09-15, 22:53	#1
The Carnivore Jun 2010 2³×3×11 Posts	The Primegrid monopoly I know I'm potentially stirring up a hornet's nest with this post, but here it goes... In all my years of watching and participating in the distributed computing world, I've never seen anything like Primegrid's takeover of other projects. Some of those takeovers were justified, but a few of them had nasty consequences that must be said, even if it's probably too late to do anything about it. Let's see some examples: Three years ago, the PCP (Prime Cullen Prime) project was taken over: http://www.mersenneforum.org/forumdisplay.php?f=79 That project searched for prime cullen primes, but instead of removing the prime cullen candidates from their Cullen search, Primegrid went ahead and tested all cullen candidates anyway, regardless of whether they were prime or not. Unfortunately, the PCP project was not the only casualty in Primegrid's ruthless march to domination. A year later, Primegrid apparently decided that it wasn't enough to destroy one project, it had to get rid of another. On September 2008, the 3*2^n-1 project fell victim to Primegrid: http://www.mersenneforum.org/forumdisplay.php?f=14 More conquests followed. The proth search project was taken over that same year: http://www.prothsearch.net/index.html Yves Gallot's Generalized Fermat Prime Search was absorbed by Primegrid soon after: http://yves.gallot.pagesperso-orange.fr/primes/index.html as was the 12121 search: http://primes.utm.edu/bios/page.php?id=640 Today, Primegrid has become more cunning and is adopting a more subtle strategy. The 17 or bust project is getting swallowed up this time. It is already hard, if not impossible, to reserve a 17 or bust sieve range outside of Primegrid: http://www.free-dc.org/forum/showthread.php?3501-Sieve-Information-getting-started and the primality testing part of the project is slowly being transferred over: http://www.free-dc.org/forum/showthread.php?25274-Now-a-predominantly-PrimeGrid-project I don't know when this will happen, but it's only a matter of time before 17 or bust also succumbs to Primegrid. Now I'm not saying that all of Primegrid's takeovers were wrong. Like I said before, some of those takeovers were justified, like when Primegrid took over Riesel Sieve since the admins of that project disappeared. But they should respect the rights and interests of other active projects, no matter how large or small they are. So don't let them take over another active effort again, even if you're not directly involved in it. If you're mainly an NPLB cruncher, join the RPS folks and speak up if Primegrid starts to test k<300 for riesel numbers. If you're mainly an RPS cruncher, join the NPLB folks and speak up if Primegrid starts to test riesel numbers for k=301-3200. And refuse to crunch for Primegrid if they ever do something like that. It may be too late to fix past wrongs, but we still can prevent future injustices.

2010-09-16, 22:12	#8
gd_barnes "Gary" May 2007 Overland Park, KS 27627₈ Posts	Contrary to what most people might expect, I am in favor of possibly "BOINCing" a couple of the bases on CRUS in the distant future; that is "asking" PrimeGrid to share the effort with us. I can foresee a day when one of our bases reaches n=4M or 5M and hasn't had any activity in 6 months or more that it might make sense. As a good example, if we are down to one final k remaining on Riesel base 6 (currently 2 k's are remaining, 1 at n=1M and 1 at n=820K) and we've searched it to n=4M with little interest shown for an extented period, that would be the time to ask PrimeGrid for help. At that point, the effort becomes virtually the equivalent of the 32^n-1 effort where one k is being searched to an extremely deep depth. In response to the "instant gratification crowd", I certainly am one. The case that I make for doing it is that it is somewhat thankless work that few people want to do. There is no glory and even low-to-mid top-5000 work usually drops off within a year. The most thankless of all is non-top-5000 work but I feel it is worthwhile to have a full and accurate listing of primes. Last fiddled with by gd_barnes on 2010-09-16 at 22:16*

2011-04-24, 20:55	#10
ellipse Nov 2010 2×3×5 Posts	Right now, Primegrid has 2391 primes on the top 5000 list, and the n=666666 primes have just started dropping off. The top 3 projects (Primegrid, RPS, and NPLB) currently have 3750 primes on that list. This means that exactly 75% of the primes on the list belongs to one of those three projects. Last fiddled with by ellipse on 2011-04-24 at 20:57

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Primegrid > GIMPS?	MooMoo2	Lounge	11	2020-08-02 20:55
ROLP at PrimeGrid	mdettweiler	Raiders of the Lost Primes	2	2010-09-17 05:57
PrimeGrid site down :(	mazadillon	Prime Sierpinski Project	2	2008-12-24 00:19
PrimeGrid dump	Sloth	Prime Sierpinski Project	4	2008-08-02 12:38
Anyone knows about primegrid	koders333	Factoring	8	2006-03-14 03:23

2010-09-15, 23:13	#2
rogue "Mark" Apr 2003 Between here and the 31·233 Posts	I am writing this to refute much of what you say. I maintain most of the projects on ProthSearch. I worked with PrimeGrid to help them get started. I also helped them get started with the Cullen and Woodall searches, which also originated on ProthSearch and had almost no remaining participants. The PCP project happens to be a subset of the Cullen Prime search. If you read this, then you will see that Rytis extends his regrets over it. I also helped them restart the GFN search. Yves Gallot has been MIA for years and that project has not moved forward since 2004 (?). There have been private efforts to extend the GFN search, but they remained private and closed to others. David Underbakke (writer of genefx64) and I worked together so that genefx64 could be used by PrimeGrid via PRPNet. David was probably the only person actively searching for GFNs. I can understand some people not being happy about PrimeGrid, but trust me, PrimeGrid has done its utmost to not step on the toes of other projects. The offer a stable and robust platform and server and it is easy to see why projects would want to get on their bandwagon. I personally believe that PrimeGrid is one of the best distributed computing projects out there because it has greatly increased visibility to the searches they are undertaking. Most of these other projects were languishing and would have died completely if it weren't for PrimeGrid. Note that PrimeGrid has avoided NPLB (No Prime Left Behind). It has also stayed away from TRP (The Riesel Problem) for many years due to the owners insisting that they would bring it back.

2010-09-16, 19:13	#4
henryzz Just call me Henry "David" Sep 2007 Liverpool (GMT/BST) 3³×227 Posts	The reason a lot of projects lose activity and/or change to primegrid is because the tests for most projects get larger and larger and take longer and longer. The only way to make any significant progess after a while is to turn to BOINC. It would surprise me if at least one of NPLB, RPS, or CRUS didn't end up asking Primegrid or BOINC for help in the next few years. I expect that the projects would all remain working here on mersenneforum as well because of the nature of the projects.

2011-03-06, 19:37	#9
Oddball May 2010 499 Posts	Just wanted to point out that Primegrid now has more primes on the top 5000 list (2207) than projects #2-20 combined (2085): http://primes.utm.edu/bios/top20.php...&by=PrimesRank That may change over the next few months when their n=666666 primes begin dropping off the list.