LLRnet and PRPnet servers for automated LLR

[Oddball]

The status of the PRPnet server here:

http://noprimeleftbehind.net:12000/all.html

has been down the past day or two, but the results found on http://nplb-gb1.no-ip.org/tps/ indicates that port 12000 has dried out.

Max, could you load n=482000-483000 into the PRPnet server?

[mdettweiler]

Originally Posted by Oddball

The status of the PRPnet server here:

http://noprimeleftbehind.net:12000/all.html

has been down the past day or two, but the results found on http://nplb-gb1.no-ip.org/tps/ indicates that port 12000 has dried out.

Max, could you load n=482000-483000 into the PRPnet server?

Indeed, it looks like Lennart got his clients switched over to the alternate server address* and cleaned out the server.

I'm loading up 482001-483000 right now. Note however that the 19 April 2011 sieve file for k=100K-200K only goes up to n=483000. Grivozavr, could you please send me a file for k=100K-200K, n=484001-484999? Thanks.

We've been working through the k=100K-200K range amazingly quickly...at this rate we'll be on to k=200K-300K very shortly. That just gave me an idea, though...perhaps we could try doing all of k=200K-400K at once this time (same n-range, twice the k-range)? We've been working quickly enough that we'd still proceed at a decently fast rate, but the admin management effort would be halved.

*For those who haven't seen the notice over at NPLB, we've been having some problems with the noprimeleftbehind.net domain lately and it isn't pointing to the right IP address. However, our old domain, nplb-gb1.no-ip.org, still works and gets to exactly the same place, so you can just replace noprimeleftbehind.net with nplb-gb1.no-ip.org in your clients or web links to get through.

[ellipse]

Originally Posted by mdettweiler

We've been working through the k=100K-200K range amazingly quickly...at this rate we'll be on to k=200K-300K very shortly. That just gave me an idea, though...perhaps we could try doing all of k=200K-400K at once this time (same n-range, twice the k-range)? We've been working quickly enough that we'd still proceed at a decently fast rate, but the admin management effort would be halved.

If we finish sieving the n=485K-490K range to 6002T, we could load n=487001-489999, k<100K into PRPnet instead of k=200K-400K. The sieve depth for the entire n=480K-490K range will be the same, so it'll make sense to test the lower k's first.

[mdettweiler]

Originally Posted by ellipse

If we finish sieving the n=485K-490K range to 6002T, we could load n=487001-489999, k<100K into PRPnet instead of k=200K-400K. The sieve depth for the entire n=480K-490K range will be the same, so it'll make sense to test the lower k's first.

Indeed. The question is, though, whether we will finish getting n=485K-490K to optimal depth in time; if Lennart continues his periodic bursts of high firepower on port 12000, we'll be needing more LLR work long before then.

[Oddball]

Originally Posted by ellipse

The "mini-rally" may be turning into something big. 1316 pairs were LLR'ed in the past hour, which will be a record 31,584 pairs/day if it's sustained.

Anyone up for an impromptu rally?

We're now completing tests at a rate of 1365 pairs/hour (32760 pairs/day if sustained)

At this rate, the PRPnet server will dry out in a few hours, so it's time to load the next range into the server. Max, could you load n=489K-490K for k=<100K if n=484K-485K from k=100K-200K is not available?

[mdettweiler]

Originally Posted by Oddball

We're now completing tests at a rate of 1365 pairs/hour (32760 pairs/day if sustained)

At this rate, the PRPnet server will dry out in a few hours, so it's time to load the next range into the server. Max, could you load n=489K-490K for k=<100K if n=484K-485K from k=100K-200K is not available?

Loading right now.

BTW @all: as recently posted over in the NPLB forum, the noprimeleftbehind.net DNS issues have been resolved. Everything should be working fine now for most people but if you're still having lingering issues, they should clear up within the next half-day or so as various DNS servers around the world update their cached info.

Also note that due to an incidental, but not directly related issue, mail sent to my max@noprimeleftbehind.net address started bouncing late yesterday. I've fixed the problem, but again, it may take a day or so for the changes to be distributed. If you have problems sending mail to me, you can send mail to max@noprimeleftbehind.net.test-google-a.com, which is independent of our DNS records and will get to the same mailbox.

[Oddball]

Max, could you load n=487001-487500 into port 12050 (LLRnet) and n=487501-489000 into port 12000 (PRPnet)? The LLRnet server has dried out, and the PRPnet server will dry out soon.

Shortly from now, gribozavr will probably be uploading the latest k<100K file in this thread:
http://www.mersenneforum.org/showthread.php?t=12260
But if the file isn't there by Monday at 10PM CDT, just go with the latest version that you have.

[mdettweiler]

Originally Posted by Oddball

Max, could you load n=487001-487500 into port 12050 (LLRnet) and n=487501-489000 into port 12000 (PRPnet)? The LLRnet server has dried out, and the PRPnet server will dry out soon.

Shortly from now, gribozavr will probably be uploading the latest k<100K file in this thread:
http://www.mersenneforum.org/showthread.php?t=12260
But if the file isn't there by Monday at 10PM CDT, just go with the latest version that you have.

Done.

It seems that with this, we have loaded the last remaining work for k<100K, n=480K-490K. However, as of right now, the 490K-495K range is hardly well-sieved; we're not quite even at 1P yet. So it would be rather inefficient to continue on to testing n>490K until sieving has progressed further. In the meantime, we do have plenty of well-sieved work in the k=100K-200K range; everything from n=484001-489999 is essentially done sieving and ready to go. I would recommend shifting the LLR focus (at least temporarily) to that range once the current server loads are done, until 490K-495K reaches a better sieve depth.

[Oddball]

Originally Posted by mdettweiler

Done.

It seems that with this, we have loaded the last remaining work for k<100K, n=480K-490K. However, as of right now, the 490K-495K range is hardly well-sieved; we're not quite even at 1P yet. So it would be rather inefficient to continue on to testing n>490K until sieving has progressed further. In the meantime, we do have plenty of well-sieved work in the k=100K-200K range; everything from n=484001-489999 is essentially done sieving and ready to go. I would recommend shifting the LLR focus (at least temporarily) to that range once the current server loads are done, until 490K-495K reaches a better sieve depth.

Sounds good!

You can also remove n=481K-483K (which was done quite a while ago) from the stats page here: http://noprimeleftbehind.net:12000/all.html

That should make the page load a bit faster.

[Oddball]

Max, could you transfer n=487100-487500 from LLRnet to PRPnet? The PRPnet server is going to dry out in the next 2-3 hours, but the LLRnet server doesn't have much activity on it now.

[mdettweiler]

Originally Posted by Oddball

Max, could you transfer n=487100-487500 from LLRnet to PRPnet? The PRPnet server is going to dry out in the next 2-3 hours, but the LLRnet server doesn't have much activity on it now.

Uh...you sure about those figures? By my reckoning (based on the ~3.5 days it took to cover 487.5K-489K), we have another 2.3 days left with the currently loaded work.

Anyway, though, no problem--at the rate we're going, we'll need the extra work in PRPnet anyway soon. I've moved n=487101-487500 from LLRnet to PRPnet. (Note that this is slightly different than the range you specified; I started it at 487101 instead of 487100 to maintain consistency with how I've generally been doing everything else.)