PRPnet beta test server

[mdettweiler]

Okay, 300K-350K is now loaded. I've also upgraded the server to v2.1.5--I think. The server still says "2.1.4" when it starts up. I'm not sure if this is a boo-boo or if it really is 2.1.4.

Anyway, though, nothing to worry about on that front. I'll see soon enough which version it really is when my client next connects to it, and if it's the wrong one it will be remedied quickly.

Edit: Ah-ha! A quick MD5 sum verification showed that I apparently was still using the same old 2.1.4 server binary. It appears the boo-boo was on my part after all.

Edit2: Okay, now we've got the real McCoy up there.

[mdettweiler]

Oh, I forgot to mention: client binaries for 2.1.5 will be posted sometime within the next few days. Possibly tomorrow.

[Flatlander]

When will we be testing PRPnet 'in anger'? Are there still outstanding issues?

[mdettweiler]

Originally Posted by Flatlander

When will we be testing PRPnet 'in anger'? Are there still outstanding issues?

No outstanding issues at this time; there are a couple things I need to do before we can start our stress test:

-Process the earlier mini-drive results from the PRPnet server and verify them with Gary's doublechecks (yes, I'm *really* late on this...)

-Wait until we officially break off work at ~n=50K from the big team sieve. Most likely this will be very soon so this shouldn't end up holding us up.

-Post PRPnet 2.1.5 client binaries for download on the NPLB web site. Currently only 2.1.2 is posted here and 2.1.5 fixes all of the little annoying things in 2.1.2. I've used 2.1.5 for quite a while locally (and on the G2000 server) and it's proven to be quite stable.

You know, come to think of it...rogue sent me an email not long ago saying that he was working on v2.2.0 now and that it would be ready soon. From what I hear, it should have a number of new changes. Of course, new features usually mean new bugs...and a whole process of patch releases. So, probably it would be best to just release 2.1.5 to you guys for now (rather than just waiting until 2.2.0 comes out) since they'll be intercompatible besides.

Anyway, just thinking out loud.

Max :smle:

[gd_barnes]

Please don't wait on n=50K-60K or whatever. We don't plan to "break off" any ranges with the sieving drive (except perhaps for n=650K-1M later on) because that was just too much hassle for such low n-ranges. We only deleted n=20K-50K because Karsten had already tested k=2000-3000/n=20K-40K before the drive started and Lennart quickly tested n=3000-3400/n=20K-50K very early in the dirve. Please don't wait for the drive to complete. Late today, I should have all of the factor ranges up to P=19T and then I'll send you the file. That would be a good time for to grab that range but we'll still leave it in the drive.

Can we please make it a priority to compare those residues so that we can do the stress test? Perhaps you can delegate that to Ian or Tim or Karsten? I could probably create an Excel spreadsheet to parse them out and compare them but I'm sure any one of you could do it in a better way that would take less time.

On the new changes in 2.2.0, let's not wait for those either unless they are correcting bugs that specifically affect us.

If the stress test works well, the stats interface is working, and Email notification of primes can be tested, we could look to put a new k=3200-3400 drive or something in that area on the server with the idea of eventually converting all drives over to new PRPnet servers if people have no objections and things are working well.


Thanks,
Gary

[mdettweiler]

Originally Posted by gd_barnes

Please don't wait on n=50K-60K or whatever. We don't plan to "break off" any ranges with the sieving drive (except perhaps for n=650K-1M later on) because that was just too much hassle for such low n-ranges. We only deleted n=20K-50K because Karsten had already tested k=2000-3000/n=20K-40K before the drive started and Lennart quickly tested n=3000-3400/n=20K-50K very early in the dirve. Please don't wait for the drive to complete. Late today, I should have all of the factor ranges up to P=19T and then I'll send you the file. That would be a good time for to grab that range but we'll still leave it in the drive.

Okay, sounds good. As soon as I get the file I'll start setting up a server for n=50K-100K.

Can we please make it a priority to compare those residues so that we can do the stress test? Perhaps you can delegate that to Ian or Tim or Karsten? I could probably create an Excel spreadsheet to parse them out and compare them but I'm sure any one of you could do it in a better way that would take less time.

Actually, you picked just the right moment to ask: I have a little bit of time today that I can spend on this stuff and decided to make a point of doing it then. So, you should (barring anything wacky) see compared residuals by the end of the day.

On the new changes in 2.2.0, let's not wait for those either unless they are correcting bugs that specifically affect us.

My thoughts exactly. I'll take care of releasing 2.1.5 today, too.

If the stress test works well, the stats interface is working, and Email notification of primes can be tested, we could look to put a new k=3200-3400 drive or something in that area on the server with the idea of eventually converting all drives over to new PRPnet servers if people have no objections and things are working well.

Yes, the stats interface and email notification works perfectly for first-pass tests going through a PRPnet server. At this time they're only imported on a daily basis (since I don't have an hourly intermediate file system set up yet for PRPnet), but things will still work exactly as they should apart from that. Everything will get in the DB, just some of it a few hours late.

The only thing that we don't have quite ready yet is a system to import doublecheck work into the DB (currently the only server doing this is PRPnet G2000). Dave is working on setting up a second parallel DB table for doublechecks so the doublechecks don't get rejected as duplicates of the first set we already have. You won't see a difference on the outside end, though this also gives us the added flexibility of being able to add doublecheck-only stats as an option later on if we'd like. Anyway, the doublecheck DB is probably not too long from being ready at this point, and in the meantime, G2000's results are just piling up in raw format. Even if that means we're a little overdue on crediting G2000's doublecheck work in the stats, it *will* eventually get in there and be credited exactly as it should.

Max

[AMDave]

Double check is in development as Max said.
Once loaded into the database, the residuals matching and pair matching is a breeze and quick too. I'd recommend getting the database to do the work.
I just hit Friday PM so I'll be getting more done on that this weekend.

[Brucifer]

Curious as to the status now of the prpnet stuff. Is it out of the test phase? What is the current version of the client now and is it located at an URL to suck it down with all the goodies in the config file pointing to the nplb server? Are any of IB's servers set up for prpnet, and/or are there any plans to do that?

[mdettweiler]

Originally Posted by Brucifer

Curious as to the status now of the prpnet stuff. Is it out of the test phase? What is the current version of the client now and is it located at an URL to suck it down with all the goodies in the config file pointing to the nplb server? Are any of IB's servers set up for prpnet, and/or are there any plans to do that?

First of all, no, PRPnet is not yet out of beta test phase. The last thing we need to do before we can bring it out of beta is to do the big stress test thing we've been talking about for a while. For a while I was holding that off until we had absolutely stable versions of both the client and server available. Version 2.2.2 has been around for a little while, and has behaved quite stably in my testing. If memory serves, the G2000 server is running the full 2.2.2 version (though there's no way for me to check right now since the server is down).

A part of the delay is due to my having dropped the ball (due to "real life" busyness) on getting v2.2.2 NPLB client packages ready for download. The most recent version of the client posted here is 2.1.2, which is missing a number of critical features and bugfixes. I'll see if I can get all that stuff ready sometime soon.

So far none of David's servers are set up with PRPnet. The plan is to start doing that once we've finished the testing phase with servers on Gary's machines (where I can more easily keep an eye on them, upgrade a server to a newer software version, etc.)

[Brucifer]

Thanx for the info Max! :-)