LLRnet server rally March 8th-9th

[gd_barnes]

The n-rate/min is virtually unchanged. I saw the temporary 600-test spike ~15-20 mins. ago where the pair I received went from n=373729 to n=374380 in a 6 min. span. Since then, I received n=374428 and n=374480 6 and 12 mins. later respectively . The n=48 and n=52 gaps in 6 mins. each is consistent with my earlier analysis of n=50 every 6 mins.

This tells me that Sheep either brought his 2 LLRing machines (all that we are aware of that he has) and cached them at 300 tests or he found an extra 4 machines for a total of 6, and cached each at 100 tests.

No problems. If we reach n=385K, I'll be happy...extremely so if we get to n=390K.


Gary

[em99010pepe]

Changed my mind, added 6 cores.

[BlisteringSheep]

Originally Posted by gd_barnes

The n-rate/min is virtually unchanged. I saw the temporary 600-test spike ~15-20 mins. ago where the pair I received went from n=373729 to n=374380 in a 6 min. span. Since then, I received n=374428 and n=374480 6 and 12 mins. later respectively . The n=48 and n=52 gaps in 6 mins. each is consistent with my earlier analysis of n=50 every 6 mins.

This tells me that Sheep either brought his 2 LLRing machines (all that we are aware of that he has) and cached them at 300 tests or he found an extra 4 machines for a total of 6, and cached each at 100 tests.

No problems. If we reach n=385K, I'll be happy...extremely so if we get to n=390K.


Gary

I finally got my local proxy working & loaded it up for my octocore. I tried to get it to cache 400 WUs, but we had a fight and it won & took 600+! I've got it set now to cache 400 and return results every 200 & prune every hour. I think. That's what I set in the llr-serverconfig.txt, but it seems to pretty much just do whatever it damned well wants!

[IronBits]

I'm finding out the llrnet servers are not that robust, or cooperative, and have a mind of their own.
You running it under M$ or *nix ? They act differently...

[mdettweiler]

Originally Posted by IronBits

I'm finding out the llrnet servers are not that robust, or cooperative, and have a mind of their own.
You running it under M$ or *nix ? They act differently...

I remember reading once on the Riesel Sieve forums that besides the basic LLRnet server included with the LLRnet client (the one we're using), they had made a whole variety of different LLRnet servers, some more complex than the others. None of those alternate LLRnet servers were released publicly, but one of the admins there said that they'd be glad to give them out to anyone who asked.

[gd_barnes]

Originally Posted by em99010pepe

Changed my mind, added 6 cores.

Ah, that explains it. I noticed that we had jumped from ~8n/min to ~10n/min in the last few hours. We still have well more than plenty of pairs in the server to cover our needs.

With you adding your cores, we have a chance at n=385K now I think.

I'm very curious to see who takes 1st place in this rally. I brought a fair amount more firepower this time around.

[IronBits]

Originally Posted by Anonymous

I remember reading once on the Riesel Sieve forums that besides the basic LLRnet server included with the LLRnet client (the one we're using), they had made a whole variety of different LLRnet servers, some more complex than the others. None of those alternate LLRnet servers were released publicly, but one of the admins there said that they'd be glad to give them out to anyone who asked.

Consider it asked. Can you make the connection and send them my way please?

[BlisteringSheep]

Originally Posted by IronBits

I'm finding out the llrnet servers are not that robust, or cooperative, and have a mind of their own.
You running it under M$ or *nix ? They act differently...

Linux. An install of ubuntu-server-7.10 I did on a little P4 with 512megs, just so that I could have another distro to play with.


I use Slackware on the firewall, Fedora on the PS3, Solaris 8 on the UltraSPARC, WinXP on the laptop & "family" PC and Gentoo on everything else (Athlon, Xeon, PowerMac, dual-boot laptop), though I've been trying (unsuccessfully) for a week to get Kubuntu on my octo. We use mostly RHEL-AS4 at work, with a couple of YDL XServes and PowerMacs.

[mdettweiler]

Originally Posted by IronBits

Consider it asked. Can you make the connection and send them my way please?

Okay, will do.

[mdettweiler]

You know, I find it kind of odd that we've only had two or three primes reported from this rally so far, given the huge range that we've tested. Maybe all those primes are just sitting around "hiding" while we're waiting for users to report them to the top-5000?

In other news, it looks like the rally won't have a problem making it to n=385K. How about we aim for n=390K then?

[gd_barnes]

Originally Posted by Anonymous

You know, I find it kind of odd that we've only had two or three primes reported from this rally so far, given the huge range that we've tested. Maybe all those primes are just sitting around "hiding" while we're waiting for users to report them to the top-5000?

It's not unusual at all at 1 PM GMT when you posted this. We had 8 primes the first rally. Even if there are no 'hiding' primes waiting to be reported, we've now had 5-6 primes this rally at 5:40 PM GMT, including 2-3 confirmed primes.

The range we've processed isn't extremely huge because it's 1/6th as many k's. We're processing a higher range than in rally 1. The odds of a prime are < 1 in 5000 vs. > 1 in 4000. Also, the # of k/n pairs should be less too, even if we have have the same firewpower.

When we have a 24-hour rally around n=500K, 1-2 primes will be close to the expectation with the same firepower...much longer testing time for each pair and less chance of each being prime.


Gary