View Single Post
Old 2008-12-31, 02:20   #6
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

100111110000002 Posts
Default

Quote:
Originally Posted by mdettweiler View Post
Okay, I like that idea!

BTW @Carlos: Thanks for the offer of running a server, though that probably won't be necessary since we already have had a G8000 server all pre-readied for this effort. However, yes, we'll definitely keep you in mind in case we need any more. (For these smaller k/n pairs, server load is somewhat higher than it is at the higher k/n pairs, so we may very well need a second server for this range.)

Hey, I just thought of an idea: how about, in addition to any LLRnet servers that we have going for this range, I run a PRPnet server for it? I'm beginning to get the hang of how this whole PRPnet thing works now, and I'm planning to write some scripts later today to ensure that PRPnet will fit into our existing stats and status-page setup. Setting up a second server for the 9th Drive would not be hard at all. Not to mention that PRPnet uses LLR 3.7.1c for numbers on power-of-2 bases.

Max

OK, let's just run n=350K-352K sieved to P=1.4T. It'd be too messy and risky to wait for higher factors from Ian and myself. I could actually complete my P=1.8T-3.4T range in 2 days if I put it on all 9 of my working quads but that would take me quite a bit of time to do because I'd have to piecemeal out the P-ranges that haven't been done yet. I also don't want to impose upon Ian to do the same with his range. It's just time-consuming to bring all the factors back together from so many cores. By just using the P=1.4T file, after everyone is done with their sieving, we'll just systematically apply all the factors P>1.4T to the entire file in an orderly fashion. I need order for my disorderly mind. lol

Running it to n=352K should work and take the pressure off of the sieving effort a little. It'll be a while before the 5000th prime is at 352K. (>4-5 weeks I think)

Max, let's hold off on the PRPnet server since we need to complete the range within a fairly short timeframe. It will take a little time for people to download everything and orient themselves to the new server software. Although that shouldn't take each person too long, I don't want to be derailed by any potential issues that there could be in the new software. Instead, let's plan on using PRPnet for n>352K after the sieving is done. As for running a server for this range, let's just use port 8000 that you've already set up. That was its intent.

I'm not going to recommend any kind of resource allocation to anyone. Everyone knows the situation on the 5000th place prime. Likely I'll put 2 quads in on the LLRing effort but may put as many as 5-6 on it depending on how popular the effort is and how quickly the 5000th place prime is moving up. In other words, I'll wait to see what the response is before shifting anything more than 2 quads. This is a good effort for people with lesser resources and I don't want the higher-resource folks to take most of the fun.

Max, have I sent you the n=200K-500K file for k=1005-2000 sieved to P=1.4T? If not, I'll send it to you right away. Just let me know. Let's plan on loading n=350K-351K in the server to get that cracking almost immediately at fairly high speed. We'll leave n=351K-352K for manual reservations in n=100 pieces.

Keep in mind that n=100 ranges will actually be ~70%+ larger than n=100 ranges on the 1st drive at the same n-range. That is about ~67% more k's plus likely 3-5% more candidates due to a sieve to P=1.4T vs. 5T.

I'll get the drive set up with manual-reservation files posted for n=351K-352K sometime in the morning between 6 and 8 AM GMT (midnight to 2 AM my time). Please plan on completing any manual reservation in < 2 weeks to make sure we stay ahead of the 5000th place prime. If an n=100 file is too big for you, you'll be able to reserve part of a file if you want. So that people have an idea of the size of the files, I'll post a # of candidates by each file.

One more thing Max...Please look on Karsten's pages and see who has reservations in the area. We already know about Thomas and I think Adam has a reservation or two in the area. I will PM them about the start of our effort. I doubt we'll impact anyone with n=350K-352K but don't want to give a bad perception if we are.

Let's move the goal for completing sieving back to Jan. 20th. I should still complete mine by Jan. 15th-16th because I'll pull quads off of port 5000 to put on LLRing n=350K-352K. The 5000th place prime shouldn't be near n=352K by the 20th but we don't want to push our luck in that regard. PrimeGrid has really been submitting the primes lately.

Thanks for being flexible everyone!

Edit: Would it make more sense to use a server by Carlos since he has offered? I'm going to Vegas from Thursday to Wednesday. If we do that, we'd just use his server for this range and plan on using port 8000 for most of the LLRnet ranges for n>352K.


Gary

Last fiddled with by gd_barnes on 2008-12-31 at 02:31
gd_barnes is offline