Welcome to the Prime Time Challenge (Reloaded)

Same song, second verse...let's hope with a different outcome. :)

PrimeGrid’s Challenge series continues with the "Prime Time Challenge (Reloaded)". The server survived its initiation but the Challenge did not. Unfortunately, there were other database related issues.

A 48 hour Challenge is being offered on PrimeGrid’s Proth Prime Search LLR application.

To participate in the Challenge, please select only the Proth Prime Search (LLR) project in your PrimeGrid preferences section. The challenge will begin 28 November 2008 18:00 UTC and end 30 November 2008 18:00 UTC. Application builds are available for Linux 32 bit and Windows 32 bit. These applications will be sent to 64 bit clients. As with all LLR application projects, there is no advantage of 64 bit over 32 bit.

NOTE: For the past several months, our primary focus has been on getting the PPS Sieve deep enough to start LLRing. While we have made great strides in this effort, we still need more help. Leading up to this Challenge, anyone with a 64 bit multiple core system is encouraged to join the Proth Prime Search sieving effort. Please see this post if you are interested as it requires us to enable the project in your PrimeGrid preferences.

ATTENTION: The primality program LLR is CPU intensive; so, it is vital to have a stable system with good cooling. It does not tolerate "even the slightest of errors." Please see this post for more details on how you can "stress test" your computer. WU’s will take ~2 minutes on fast/newer computers and 4+ minutes on slower/older computers. If your computer is highly overclocked, please consider "stress testing" it. Sieving is an excellent alternative for computers that are not able to LLR. :)

Please, please, please make sure your machines are up to the task.

Time zone converter:

The World Clock - Time Zone Converter

Scoring Information

Scores will be kept for individuals and teams. Only work units issued AFTER 28 November 2008 18:00 UTC and received BEFORE 30 November 2008 18:00 UTC will be considered for credit. We will use the "prime score" method which is based on the n value (k*b^n-1) to score the challenge. The only difference is that the primary and double checker of a WU will receive the same score.

Therefore, each completed WU will earn a unique score based on its n value. The higher the n, the higher the score. This is different than BOINC cobblestones! A quorum of 2 is NOT needed to award Challenge score - i.e. no wingman. Therefore, each returned result will earn a Challenge score. Please note that if the result is eventually declared invalid, the score will be removed.

For details on how the score is calculated, please see this thread.

About the Proth Prime Search

The Proth Prime Search is done in collaboration with the Proth Search project. This search looks for primes in the form of k*2^n+1. With the condition 2^n > k, these are often called Proth primes. This project also has the added bonus of possibly finding factors of "classical" Fermat numbers or Generalized Fermat numbers. As this requires PrimeFormGW (PFGW) (a primality-testing program), once PrimeGrid finds a prime, it is then manually tested outside of BOINC for divisibility.

Our initial goal is be to double check all previous work up to n=500K for odd k<1200 and to fill in any gaps that were missed. PG LLRNet searched up to n=200,000 and found several missed primes in previously searched ranges. Although primes this small do not make it into the Top 5000 Primes database, the work is still important as it may lead to new factors for "classical" Fermat numbers or Generalized Fermat numbers. While there are many GFN factors, currently there are only about 270 "classical" Fermat number factors known.

Once the initial goal is reached, we'll turn our focus to smaller k values and higher n values. For example, k<32 complete to n=2M, k<64 complete to n=1M and so on. Primes found in these ranges will definitely make it into the Top 5000 Primes database.

For more information about "Proth" primes, please visit these links:

• Proth Prime - The Prime Glossary
  
• Proth Prime - Wolfram MathWorld
  
• Proth Number - Wikipedia

About Proth Search

The Proth Search project was established in 1998 by Ray Ballinger and Wilfrid Keller to coordinate a distributed effort to find Proth primes (primes of the form k*2^n+1) for k < 300. Ray was interested in finding primes while Wilfrid was interested in finding divisors of Fermat number. Since that time it has expanded to include k < 1200. Mark Rodenkirch (aka rogue) has been helping Ray keep the website up to date for the past few years.

Early in 2008, PrimeGrid and Proth Search teamed up to provide a software managed distributed effort to the search. Although it might appear that PrimeGrid is duplicating some of the Proth Search effort by re-doing some ranges, few ranges on Proth Search were ever double-checked. This has resulted in PrimeGrid finding primes that were missed by previous searchers. By the end of 2008, all new primes found by PrimeGrid will be eligible for inclusion in Chris Caldwell's Prime Pages Top 5000. Sometime in 2009, it is expected that over 90% of the tests handed out by PrimeGrids will be numbers that have never been tested.

PrimeGrid intends to continue the search for indefinitely for Proth primes.

What is LLR?

The Lucas-Lehmer-Riesel (LLR) test is a primality test for numbers of the form N = k*2^n − 1, with 2^n > k. Also, LLR is a program developed by Jean Penné that can run the LLR-tests. It includes the Proth test to perform +1 tests and PRP to test non base 2 numbers. See also:

• Lucas-Lehmer-Riesel test (WIKI)
  
• Download LLR by Jean Penné
  

(Edouard Lucas: 1842-1891, Derrick H. Lehmer: 1905-1991, Hans Riesel: born 1929).
____________

We would like to pre-stress the server on 27 Nov 2008 at 17:00 UTC. We will be monitoring to see if the previous Challenge related issues present themselves again.

If you are interested in helping out, please select Proth Prime Search (LLR) in your PrimeGrid preferences and hit the server at the prescribed time. We are hoping for a good mixture of small and large cache hits (EDIT: a few minutes to a few hours).

Thank you for your help!
____________

Will you send workunits at 17:00 UTC ??
I´am worry about get a large number of workunits, which doesn´t count for the challenge starting at 18:00 UTC
____________
Sysadm@Nbg
my current lucky number: 113856050^65536 + 1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/

That's 17:00 the day before the challenge starts.
____________
There are only 10 kinds of people - those who understand binary and those who don't

That's 17:00 the day before the challenge starts.

That's 17:00 the day before the challenge starts.

Uups I didn´t read exactly enough
Sorry ...

There are 25 hours to complete the pre-stressing workunits

I'm in doubt. I think that i'm not in until you describe whats happened at first time this challenge starts. I'm curious admin, so you don't scare me with words like raid, mysql, etc. And I think there's a lot of tech guys who also curious as well. so please tell us WHAT'S HAPPENED AT FIRST TIME.
____________
wbr, Me. Dead J. Dona

If we knew exactly, we'd not need a pre-stress test :)

Basically, the servers were not fully loaded (unix load level below 2 on WWW server - 2 cores available; Windows load level below 20% on DB server - 8 cores availabe), but access to the website (and work issuing) stopped. This might be because:
a) Network issues between DB and WWW servers;
b) MySQL stalling while handling some specific query;
c) Apache webserver stopping responding;
d) BOINC bug, stalling work send.

Sadly, I wasn't around when it happened, so with the pre-stress test I hope to pinpoint the exact cause out of these. If I had to bet, it's either b) or c). a) is unlikely (nothing in the logs about lost connections). I don't know for sure, and cannot attempt a fix until I know.
____________

Thanks for info, Rytis.
Can you give us numbers on load like db querries per sec, connections per sec on network?

As for pre-challenge stress-test, I may have couple dozens of cores...
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

The load balancer/firewall of the data center the server is installed in is out of consideration?
As I understood the server was hammered with requests - at work our web application security software (WASS) would have banned those clients temporarily on a per-IP-basis.
This would amount to following scenario:
- clients are asking for work
- clients get work
- work obtained by client doesn't satisfy cache-settings of project
- clients asking for work
- clients getting response that work-quota is reached
- clients go to sleep for 60 secondes
- clients are asking over and over for work and getting response that work quota is reached
- WASS is blocking those clients due to DDOS-pattern
- no more work issued for these clients
- website also not accessible for these clients

just my 2 cents...

If the data center does not have WASS - which I can't quite imagine - that scenario is out of question.

kind regards

p.s.: I will try to do my best @ 27.th and the following two days...
____________

@ roadrunner_gs - AFIAK, clients go to sleep for longer and longer periods when no work available.

Blocking website on per-IP address is not a wise idea. Users would not be able to look at forum to see what's going on, report problems etc. Futher more, clients would not be able to upload and report work already done. If they would be behind NAT, a whole farm might get blocked because of a single client going nuts...

The issue of network load is a good point nevertheless.
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

I'm sure you read all logs, and if u say u don't know why this happend, this is a UNKNOWN issue. Bad. Hope you catch it at 27. If something should happen let it be at 27 not 28.
Does BOINC server version is up to date?
____________
wbr, Me. Dead J. Dona

Ten minutes of stressing ...
Meanwhile sporadig message: no work sent

Took 26 minutes to get work for all 21 machines.

I switched my desktop today to PPS-LLR. But there is something tricky for me. Normally it takes around 2:11 minutes to finish one WU. And from time to time there are some WUs that take 2:50 or longer. Is this normal?

Edit says: Just looked on the laptop. Same thing there. Most WUs around 380 seconds, some over 500 seconds.

Just wondered, that this was not at all WUs.

The pre-stress on the server to the Prime Time Challenge (Reloaded) has shown a substantial improvement. A few more adjustments should make things even better. Therefore, the Challenge is on for 28 Nov – 30 Nov.

Changes to Apache settings helped alleviate the problems. Additionally, we will be making adjustments to the available WU buffer (what you see on the front page), max WU cache (how much a single core can download at one request), and project WU share (how many WU's BOINC has available for the PPS LLR project).

These changes will only affect the start. Once everyone has work, all settings will return to normal.

Thank you to everyone who participated in the pre-stress. If you wish to stay on till the start of the Challenge, please remember to lower your cache before 18:00 UTC tomorrow so you can optimize your Challenge score. :)

Best of Luck in the Challenge and in finding primes!
____________

The computational time is increasing on a regular basis. Two days ago my Laptop (the absolutely slowest Core Duo you can find) finished PPS-LLRs in 600 seconds. Then 800 seconds. Now the workunits take either 1000 or 1300 seconds. The computation time is increasing quite fast, considering the doubling within 2 days. During the challenge with more people crunching explicitly for this subproject, do I have to expect another factor of 4 or 5 over the weekend?

Lets see how many workunits I can crunch with my Quad Core. But I am pretty sure that it is impossible to reach any reasonable challenge score (with reasonable I mean not zero) with my two processors/6 cores, although running them 24/7.

He doesn't, I took over and adjusted it the way it's running fine now :)
____________

It's not been 100% flawless, but I'm pretty happy with the way this challenge started compared to many previous ones. All my hosts got work without having to wait for too long, are gradually filling their caches, and the website has been available almost all the time. Well done chaps!
____________

Great job guys, thank you :)
____________

as said inside the chat: this time challenge is running
thank you so far
____________
Sysadm@Nbg
my current lucky number: 113856050^65536 + 1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/

Just attached for the challenge a few minutes ago, and am getting the

11/28/2008 13:19:27|http://www.primegrid.com/|Sending scheduler request: Project initialization. Requesting 1 seconds of work, reporting 0 completed tasks
11/28/2008 13:19:37|PrimeGrid|Scheduler request succeeded: got 0 new tasks
11/28/2008 13:19:37|PrimeGrid|Message from server: No work sent
11/28/2008 13:19:37|PrimeGrid|Message from server: No work is available for PPS LLR
11/28/2008 13:19:37|PrimeGrid|Message from server: No work available for the applications you have selected. Please check your settings on the web site.


Is there going to be work available for this challenge - enough to run it 24/7?

Just attached for the challenge a few minutes ago, and am getting the

11/28/2008 13:19:27|http://www.primegrid.com/|Sending scheduler request: Project initialization. Requesting 1 seconds of work, reporting 0 completed tasks
11/28/2008 13:19:37|PrimeGrid|Scheduler request succeeded: got 0 new tasks
11/28/2008 13:19:37|PrimeGrid|Message from server: No work sent
11/28/2008 13:19:37|PrimeGrid|Message from server: No work is available for PPS LLR
11/28/2008 13:19:37|PrimeGrid|Message from server: No work available for the applications you have selected. Please check your settings on the web site.


Is there going to be work available for this challenge - enough to run it 24/7?

Thanks for all the feedback. We will continue to make adjustments and increase max WU cache...currently at 119 WU's.

BTW, we just passed n=400K. The Challenge started at n~393K. Great work everyone!!!
____________

validator shows nearly 100k pending.

might need a kick or 2. ;)

29/11/2008 6:18:59 PM|PrimeGrid|Message from server: Completed result pps_llr_24776286_0 refused: result already reported as success


Whats this about-- I now have about 70 of these messages across 3 hosts-- anyone else seeing the same???--- Are there duplicate WU's being sent out???-- I can not find any indication in messages where I have double reported.

29/11/2008 6:18:59 PM|PrimeGrid|Message from server: Completed result pps_llr_24776286_0 refused: result already reported as success


Whats this about-- I now have about 70 of these messages across 3 hosts-- anyone else seeing the same???--- Are there duplicate WU's being sent out???-- I can not find any indication in messages where I have double reported.

One of my quads has ~130 of these since challenge started. It is the one that failed to report larger portion of completed results. When it succeded, many results were already checked and WUs closed - hence the message I think.
Another host has only ~10 of them.
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

[quote]

So everything is ok, no work lost.

I see the PPS WU's are stretching out.

Duo Core --4' 40" ( 2' 30" 2 weeks ago)

AMD (1 CPU) --7' 50"

P4-- 10' 30" (poor ole' P4-- should retire it)

(11am UTC)

How are other hosts comparing--

Numbers are getting bigger!!!

Pete, it was stated very clearly that "Only work units issued AFTER 28 November 2008 18:00 UTC and received BEFORE 30 November 2008 18:00 UTC will be considered for credit." - see John's first post in this thread.
You don't need to abort units that don't fit to the challenge. Just suspend them (if not in danger of miss deadline) and resume after challenge is done (or when you go out of work).

I agree it may be discouraging if you do not have permanent internet connection to enter such a challenge as this. But it's a challenge, isn't it?

P.S. I have 26 more hosts not attached to PG and still in hesisation to enter challenge with them because I also have some concerns....
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

Honza-- Good comments-- Comiserations to Cruncher Pete--- HOWEVER be careful about suspending old WU's-- Your cache will still recognise this-- I know because I have one in one host-- to keep this host in the challenge (my best PC) I have to manually get work every 4 hrs or so---

I would like to congradulate PG-- despite my problems things have gone great so far--

Pete-- From a fellow Aussie-- Sh*?t happens mate-- The first challenge (aborted) also had me buggered-- I ended up suspending tasks and emptying my cache to try to get into the race (I usually crunch the longer WU's)-- My RAC still hasn't recovered from the first fiasco--- But I'm here again for the next round-- Read the threads mate -- all the info is there-- Careful about suspending tasks though-- see my thread

Seems like the recession is affecting prime work too...

The server is sending me "pink slips" saying, "Message from server: No work is available for PPS LLR".

The main page also says, "Available: Proth Prime Search (LLR) 0".

I didn't have a very big cache; I'm going to have to start doing sieve if I don't get more work very soon.

I have about at minimum 10 hours of work on my three fat clients (C2D T8300, C2D E6550, Xeon 5405, Pentium M 2000), i am not frightened at the moment... ;)

For those whose clients are not sending work, are your preferences set to allow network use all the time?

Meanwhile, my fast machine's out of work, trying to get a sieve when it can (those are being reported as unavailable sometimes too), and my slow machine will run out in a little over 45 minutes. (3 WU - it's really slow!)

Also running out of work on multiple boxes, I've got a 0.5 day bunker and 20 cores are already idling...

2 out of 3 of my computers have work. The 1 computer out of work is the one that caches only 3 WUs at a time.
____________

Seems Rytis lovingly kicked the server. Now there are some PPS LLR WUs. Thanks.

I wonder why admins just doesn't generate say 1kk wus to ensure that everyone get their work?
____________
wbr, Me. Dead J. Dona

I have had over 500 kicked because they were already reported as successes or resent because lost. have you also noticed the teams are updating scores but participants are not. Granted just getting the bugs out but it's always cool to see how your doing. I did read somewhere that some of the wu 's would be duplicates but they said it would be minimal.

29/11/2008 6:18:59 PM|PrimeGrid|Message from server: Completed result pps_llr_24776286_0 refused: result already reported as success


Whats this about-- I now have about 70 of these messages across 3 hosts-- anyone else seeing the same???--- Are there duplicate WU's being sent out???-- I can not find any indication in messages where I have double reported.

This happens, if the server load is that high, so that your client's connection times out. The server still gets your results and mark them as reported, but it can't tell your client to delete them from your list. Your client will retry to report the tasks, but the server already marked them as reported in the database and returns that message.

So everything is ok, no work lost.

The timeouts are happening simply because there are too many connections to the server. While they are a nuisance, they are expected with all the activity that's hitting PrimeGrid's server right now.

Hopefully everyone has at least 20-30 minutes cache and therefore, can handle a time out or two...or any of the other load related messages.

If this was expected behavior, then a notice to up one's cache (at least to more than the .25 days default) would have been nice. Live and learn I guess...

I know...and I am not complaining. Just mad at myself a bit for not thinking about it ahead of time (too much of real life stuff going on for the holiday).

The 3 hour outage caused a side effect that has been problematic for me. With the shorter cache settings (a good default for those of us on multi-projects normally), the machines ran dry, and then with the increasing BOINC server backoff times, are now sitting running the other projects until another timed update (despite the drastic increase in resource share that I normally implement that essentially devotes the machines to PrimeGrid during the challenges).

Perhaps for future challenges, it might be a good idea to recommend a cache setting of at least 50% of the challenge duration?

Edit: The machines affected are remote, unfortunately.
____________
141941*2^4299438-1 is prime!

Well.. there was a post somwhere I read telling us to increase the cache once the challenge was successfully under way---

It would be very helpful if the PG guys put out a list of do's & don'ts & recommendations prior to a challenge for us poor fools who don't know a lot about how PG & BOINC works--

I suspended a big WU (90%) just before the start-- I had no idea this would prevent this host from requesting work & apparently reporting.
This is stuff I need to know-- Luckily I watched the start and figured out the issue after about 15min--

Following the posts I see a lot of others were caught by this issue--

Even a page of script somwhere explaining the various functions etc of the GUI would help-- All you need to know seems to be in the message posts but it gets pretty tedious searching all these.

It would be very helpful if the PG guys put out a list of do's & don'ts & recommendations prior to a challenge for us poor fools who don't know a lot about how PG & BOINC works--

no update for individual stats since yesterday evening. :(

no stats - no fun - no race...

This happens, if the server load is that high, so that your client's connection times out. The server still gets your results and mark them as reported, but it can't tell your client to delete them from your list. Your client will retry to report the tasks, but the server already marked them as reported in the database and returns that message.

So everything is ok, no work lost.

30/11/2008 11:42:40|PrimeGrid|Message from server: Resent lost result pps_llr_25190407_0

no update for individual stats since yesterday evening. :(

no stats - no fun - no race...

[B@H] Clooney - BOINC knows more than one project. You should have set another project(s) with a low share as a backup. Or should have used better BOINC cores like Boinc studio, which has ability to set backup projects (if no work for main project, start backup projects and get as little work just to avoid cores being idle).
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186