mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   No Prime Left Behind (https://www.mersenneforum.org/forumdisplay.php?f=82)
-   -   LLRnet server rally port 300 May 3rd-4th (https://www.mersenneforum.org/showthread.php?t=10256)

gd_barnes 2008-05-02 20:28

LLRnet server rally port 300 May 3rd-4th
 
We would like to run a 24-hour LLRnet server rally on port 300 from May 3rd at 7 PM GMT to May 4th at 7 PM GMT.

This would be noon Sat.-noon Sun. PDT, 2 PM-2PM CDT, or 3 PM-3 PM EDT in the U.S.

Here is the server info.:
server = "nplb.rieselprime.org"
port = 300

General info. on setting up a connection to an LLRnet server can be found [URL="http://www.mersenneforum.org/showthread.php?t=9959"]here[/URL]. The above info. should be put in the llr-clientconfig.txt file.

Everyone bring all of your firepower to the rally as we see how fast we can bring drive 1 up to n=425K! :smile:




Gary

Mini-Geek 2008-05-02 21:16

Ready to rumble. :smile:
The rally starts right when my brother's birthday party does, so we'll probably be doing some gaming (GH3 with custom songs) on it and so I'll be slowed for a while...I'll leave it on LLRNet work long enough to get in 24 CPU hours per core to compensate.

gd_barnes 2008-05-02 21:22

I'm bringing 2 quads plus at least 2 more cores to the rally so all of you heavy-hitters, bring it on! :smile:

Edit: Which reminds me, I need to get those Linux machines loaded up with LLRnet here shortly. Ah...more Linux learning curve. Fortunately Anon was kind enough to forward me some detailed instructions while I'm gone.

mdettweiler 2008-05-03 01:57

Gary, you need to make a note in the first post of this thread that the server is "nplb.rieselprime.org", since if people follow the instructions you linked to with port 300, they'll be trying to access Carlos's now-defunct port 300 server.

In other news, I'll bring two cores of a Core 2 Duo 2.2Ghz to the rally. :smile:

mdettweiler 2008-05-03 02:16

Oh, one more thing: I'll get some more knpairs put together for Adam to add to the port 300 server, since we'll want it to be fully loaded for the rally. :smile:

gd_barnes 2008-05-03 02:35

[quote=Anonymous;132629]Gary, you need to make a note in the first post of this thread that the server is "nplb.rieselprime.org", since if people follow the instructions you linked to with port 300, they'll be trying to access Carlos's now-defunct port 300 server.

In other news, I'll bring two cores of a Core 2 Duo 2.2Ghz to the rally. :smile:[/quote]

Specific server info. has now been added.

AES 2008-05-03 03:54

Please keep in mind that this is a Windows server and if a proxy is used (proxyMasterAddress), it will eventually stop releasing sockets and seize up.

Also, if you would like an email notification of new primes found, please send the email address and user name to Anon or AES.

mdettweiler 2008-05-03 04:11

[quote=AES;132638]Please keep in mind that this is a Windows server and if a proxy is used (proxyMasterAddress), it will eventually stop releasing sockets and seize up.

Also, if you would like an email notification of new primes found, please send the email address and user name to Anon or AES.[/quote]
So you're saying that personal proxies bog down the Windows servers? In that case, should users try to avoid using personal proxies on your server? It would mean more load on your network, but that's better than having the server lock up. :smile:

What you might want to consider is to install Virtual PC, VMware Player or some other free virtualization software on your Windows server machine, run a Linux/Unix variant in a VM, and run the LLRnet server from there. IronBits has reported that the Linux server is without a lot of the Windows server's idiosyncracies, so it might be quite beneficial to run it instead.

AES 2008-05-03 05:01

It doesn't really bog the server down or cause excess cpu/disk usage, it just causes it to stop releasing sockets and seize up for some reason. We've observed this on all Win32 servers we've set up. I don't know if llrnet clients alone will cause this.

The network traffic is negligible either way.

em99010pepe 2008-05-03 13:09

[quote=gd_barnes;132613]I'm bringing 2 quads plus at least 2 more cores to the rally so all of you heavy-hitters, bring it on! :smile:
[/quote]

I think you should bring everything....added 6 cores to the fun.

Carlos

gd_barnes 2008-05-03 14:10

[quote=AES;132641]It doesn't really bog the server down or cause excess cpu/disk usage, it just causes it to stop releasing sockets and seize up for some reason. We've observed this on all Win32 servers we've set up. I don't know if llrnet clients alone will cause this.

The network traffic is negligible either way.[/quote]

Adam,

How is your server currently responding to things? I've had the full gammit of my machines that I'll have for the rally on port 300 for about 8 hours now. I'm not sure if Carlos has moved his 6 cores to it yet.

Do we need to encourage everyone to move all their machines at least an hour before the rally to make sure your server can handle it?

I'm going back to bed now...will be back on in ~3 hours.


Gary

AES 2008-05-03 14:24

no problem so far

mdettweiler 2008-05-03 16:48

Both cores of my primary machine have been added to port 300...I'll probably be able to get a P3 1Ghz on it a little later, too. :smile:

em99010pepe 2008-05-03 17:06

Adam,

XtremeSystems is on a move, could you update the team stats? Stay alert for AMDUsers, Anandtech and Ars members too. Thanks.

Carlos

mdettweiler 2008-05-03 17:21

[quote=em99010pepe;132663]Adam,

XtremeSystems is on a move, could you update the team stats? Stay alert for AMDUsers, Anandtech and Ars members too. Thanks.

Carlos[/quote]
Okay, I just added "XtremeSystems_glennpat" (the only XtremeSystems person who has contributed to the server yet, it would appear) as a member of team XtremeSystems. Adam, just to let you know, I marked his email address as "null@spambob.org" (a disposable email address that redirects all incoming mail to /dev/null immediately upon receipt), since I don't know his address and it wouldn't let me leave the address field blank.

I still don't see XtremeSystems in the team stats yet, though, despite having added a member to it on the admin page...Adam, is there something else that has to be done in order for XtremeSystems to show up in the stats?

mdettweiler 2008-05-03 17:27

[quote=Anonymous;132664]Okay, I just added "XtremeSystems_glennpat" (the only XtremeSystems person who has contributed to the server yet, it would appear) as a member of team XtremeSystems. Adam, just to let you know, I marked his email address as "null@spambob.org" (a disposable email address that redirects all incoming mail to /dev/null immediately upon receipt), since I don't know his address and it wouldn't let me leave the address field blank.

I still don't see XtremeSystems in the team stats yet, though, despite having added a member to it on the admin page...Adam, is there something else that has to be done in order for XtremeSystems to show up in the stats?[/quote]
Ah, now XtremeSystems shows up in the team stats. Okay, now that's all set! :smile:

em99010pepe 2008-05-03 17:39

Don't worry about his email, I'll contact him via XS forum.

mdettweiler 2008-05-03 17:45

[quote=em99010pepe;132667]Don't worry about his email, I'll contact him via XS forum.[/quote]
Thanks. In that case, would you rather that I set his email address to "celppinho+glennpat AT gmail DOT com", so that you'll receive his prime notifications and can then notify him? (The +glennpat in the email address will mean that you can sort all his emails into a separate label if you want.)

gd_barnes 2008-05-03 17:59

I'm back on now. Is everything still looking good? It sounds like we may have some serious firepower going after this one! :fusion:

Xyzzy is being his funny self again...my avitar is upside down. lmao

Mini-Geek 2008-05-03 18:14

[quote=gd_barnes;132670]I'm back on now. Is everything still looking good? It sounds like we may have some serious firepower going after this one! :fusion:

Xyzzy is being his funny self again...my avitar is upside down. lmao[/quote]
I added my two cores about an hour ago. 25 results so far. I noticed your avatar earlier...maybe the snake scared the gerbil into flipping your avatar. Yeah, this is going to be a good rally. :smile:

Lennart 2008-05-03 23:37

I am on from now
 
Hello

I help you out on the rally! Have 20 core on now.
Tell me if i shall hold of.

/Lennart

mdettweiler 2008-05-04 01:30

[quote=Lennart;132685]Hello

I help you out on the rally! Have 20 core on now.
Tell me if i shall hold of.

/Lennart[/quote]
Wow, 20 cores! :shock: I say keep it up! :grin: :grin: :grin:

Lennart 2008-05-04 01:35

[quote=Anonymous;132688]Wow, 20 cores! :shock: I say keep it up! :grin: :grin: :grin:[/quote]


Ok I have now 30 core on the server and have to stay there becuse i cant take the other 13 quad off sieving.

/Lennart

mdettweiler 2008-05-04 01:37

[quote=Lennart;132689]Ok I have now 30 core on the server and have to stay there becuse i cant take the other 13 quad off sieving.

/Lennart[/quote]
No problem--30 cores in and of itself is still a HUGE help! :w00t:

mdettweiler 2008-05-04 01:40

I just noticed that there is a user called "linux" returning results to the LLRnet server. That sounds like it could be part of some sort of internal naming system--does anyone know if this is a mistake or not? (If necessary, AES can merge it with the user's correct name after the rally--he's done that in the past with other people.)

Lennart 2008-05-04 03:37

[quote=Anonymous;132691]I just noticed that there is a user called "linux" returning results to the LLRnet server. That sounds like it could be part of some sort of internal naming system--does anyone know if this is a mistake or not? (If necessary, AES can merge it with the user's correct name after the rally--he's done that in the past with other people.)[/quote]

It was me!! Forget to set my call on one It shall be sm5ymt

/lennart

mdettweiler 2008-05-04 04:16

[quote=Lennart;132696]It was me!! Forget to set my call on one It shall be sm5ymt

/lennart[/quote]
Okay, no problem, I'm sure AES can fix it up for you once the rally's over and things have "settled down" a bit. :smile:

mdettweiler 2008-05-04 04:53

I just noticed that two users, one named "AMDUsers_Brucifer" and the other named just "Brucifer" showed up in the LLRnet stats--I'm assuming that the latter was an error since it doesn't have the team name in it. I've added both to the AMD Users team for the time being; after the rally Adam can merge the stats for "Brucifer" into "AMDUsers_Brucifer". :smile:

Brucifer 2008-05-04 05:32

[QUOTE=Anonymous;132698]I just noticed that two users, one named "AMDUsers_Brucifer" and the other named just "Brucifer" showed up in the LLRnet stats--I'm assuming that the latter was an error since it doesn't have the team name in it. I've added both to the AMD Users team for the time being; after the rally Adam can merge the stats for "Brucifer" into "AMDUsers_Brucifer". :smile:[/QUOTE]

yeah, I was experimenting. They should both be under AMDUsers_Brucifer. :)

Having an issue though on 64-bit linux systems. Is the linux llrnet client statically compiled or not? Runs okay on 32-bit linux, except slackware v12.0 which is 32-bit. So it's a library issue, as earlier 32-slack systems run it fine. Any comments?

mdettweiler 2008-05-04 05:52

[quote=Brucifer;132699]yeah, I was experimenting. They should both be under AMDUsers_Brucifer. :)

Having an issue though on 64-bit linux systems. Is the linux llrnet client statically compiled or not? Runs okay on 32-bit linux, except slackware v12.0 which is 32-bit. So it's a library issue, as earlier 32-slack systems run it fine. Any comments?[/quote]
I think it's dynamically linked, though if you want to try compiling it as a statically linked application you can get the source code [URL="http://stats.rieselsieve.com/download/llrnet-sources.tgz"]here[/URL]. (Instructions for compiling are included. It's actually pretty easy--I've done it a couple times myself--though I imagine it would be a little more complicated if you wanted to have it statically linked.)

Edit: Oh, forgot to ask, what exactly is the problem you're having on Slackware 12.0? Does it just not even start at all? Or does it have problems connecting to the server? If it's the latter then that's a known issue with DNS resolution that shows up on certain Linux distros--a recompile should fix it. (Last I heard, the affected distros were Debian and Redhat based distros--which includes me since I'm on Ubuntu--though I vaguely remember someone mentioning problems on Slackware. If you want I can send you my compiled version, compiled on a Ubuntu system running 7.10 at the time.)

gd_barnes 2008-05-04 07:02

I am thrilled at all of the new people joining the rally!

Welcome Lennart and Brucifer to the effort! :smile:


Gary

Brucifer 2008-05-04 07:05

[QUOTE=Anonymous;132701]
Edit: Oh, forgot to ask, what exactly is the problem you're having on Slackware 12.0? Does it just not even start at all? Or does it have problems connecting to the server? If it's the latter then that's a known issue with DNS resolution that shows up on certain Linux distros--a recompile should fix it. (Last I heard, the affected distros were Debian and Redhat based distros--which includes me since I'm on Ubuntu--though I vaguely remember someone mentioning problems on Slackware. If you want I can send you my compiled version, compiled on a Ubuntu system running 7.10 at the time.)[/QUOTE]

yep, boils down to resolution, it's happening for me on mepis-64, opensuse 10.2, 10.2, 10.3, and slack 12.0. Something with the way they are handling the "localhost" issue within. Slack 10.x and earlier aren't having any issues.

You have an url for your recompiled version where I could suck it down?

Also, is this server setup going to be good after this push is over Sunday?

gd_barnes 2008-05-04 07:11

Adam,

We have some serious firepower on this rally! Your server seems to be handling the load very well. Let us know if you encounter any issues.


Gary

Brucifer 2008-05-04 08:09

[QUOTE=Anonymous;132701]I think it's dynamically linked, though if you want to try compiling it as a statically linked application you can get the source code [URL="http://stats.rieselsieve.com/download/llrnet-sources.tgz"]here[/URL]. (Instructions for compiling are included. It's actually pretty easy--I've done it a couple times myself--though I imagine it would be a little more complicated if you wanted to have it statically linked.)[/QUOTE]

Well the recompile worked on the slack 12.0. However my other systems are running 64-bit linux, and when trying to recompile the 32-bit code on a 64-bit o/s, make errors out as a 32-bit math operation doesn't jive with the 64-bit world. So is there a command to force a 32-bit compile using a 64-bit make/gcc/etc??? It doesn't show anything when reading the help on the 64-bit system.

As for copying the slack recompile over to the opensuse/mepis, I get a fatal kernel error as the kernel is too old..... sigh...

Brucifer 2008-05-04 08:10

[QUOTE=gd_barnes;132704]I am thrilled at all of the new people joining the rally!

Welcome Lennart and Brucifer to the effort! :smile:


Gary[/QUOTE]


Thankyou for the welcome. If I can get the compile thing straightened out I can bring a lot more power on board. :)

Mini-Geek 2008-05-04 11:30

[quote=Brucifer;132705]Also, is this server setup going to be good after this push is over Sunday?[/quote]
Yes. This server is one of our permanent NPLB servers. It currently has about 69000 more pairs reserved.

mdettweiler 2008-05-04 13:57

[quote=Brucifer;132712]Well the recompile worked on the slack 12.0. However my other systems are running 64-bit linux, and when trying to recompile the 32-bit code on a 64-bit o/s, make errors out as a 32-bit math operation doesn't jive with the 64-bit world. So is there a command to force a 32-bit compile using a 64-bit make/gcc/etc??? It doesn't show anything when reading the help on the 64-bit system.

As for copying the slack recompile over to the opensuse/mepis, I get a fatal kernel error as the kernel is too old..... sigh...[/quote]
Okay, here you go:

[URL]http://bugmesticky.googlepages.com/llrnet.linux.sr5.tgz[/URL]

This is actually the SR5 version of LLRnet, which works on non-base 2 numbers, too (the stock client only does base 2), so it will work with our Conjectures 'R Us LLRnet servers, too. I have both the SR5 and stock versions compiled, but since neither has the GUI anyway (I don't have the GUI libraries on my system), there's no difference except that the SR5 version is a little more flexible. :wink: So I figured I'd just give you the link to that, especially seeing as how I had it already uploaded to my web space, anyway. :smile:

Edit: Oh, about your 64-bit system, to tell you the truth, I'm not too experienced with this compiling stuff myself, so I'm afraid I won't be of much help. However, my compiled version I linked to above should work on your 64-bit systems. :smile:

em99010pepe 2008-05-04 14:55

1 Attachment(s)
Interesting chart:

[url]http://nplb.rieselprime.org/stats/stats_by_date.php[/url]

gd_barnes 2008-05-04 19:18

[quote=em99010pepe;132727]Interesting chart:

[URL]http://nplb.rieselprime.org/stats/stats_by_date.php[/URL][/quote]

Most impressive! What a tremendous rally we have had!! Thanks to everyone who contributed. Karsten and Anon, I think you can start compiling the results and stats for the rally.

In the mean time, I'll be over in the CRUS forum continuing the updating of web pages there. There's still quite a few updates for bases 3 and 19 and, I think, one other base. All 3 bases have many primes found and remaining k's that need to be shown.


Gary

gd_barnes 2008-05-04 19:26

[quote=Brucifer;132705]
Also, is this server setup going to be good after this push is over Sunday?[/quote]

I'll add to what Anon stated. We have 2 permanent servers. One on port 300 that you have been running for the rally and one on port 5000. The port 5000 server is at a higher n-range right now but the primes that will be found from it have a higher chance of having already been found. That is we're effectively double-checking SOME ranges that have presumably already been searched; not all double-checking but some.

The info. for the 2 permanent servers is in [URL="http://www.mersenneforum.org/showthread.php?t=10042"]this thread[/URL]. The only things you would have to change to run port 5000 are the server name and port.


Gary

Brucifer 2008-05-04 19:42

[QUOTE=Anonymous;132725]Okay, here you go:

[URL]http://bugmesticky.googlepages.com/llrnet.linux.sr5.tgz[/URL]

So I figured I'd just give you the link to that, especially seeing as how I had it already uploaded to my web space, anyway. :smile:
[/QUOTE]

Well it worked on one of my mepis-64 systems. On another identicle system it wouldn't run. Go figure. Can't compile the 32-bit thing on that one either as I don't know how to do the 32-bit compiling on 64-bit systems without getting the errorcode.

But one more dual system is better than none. :)

kar_bon 2008-05-04 23:54

the scoring and some data about the third NPLB-race can be found here:
[url]http://www.rieselprime.org/nplb/NPLB_Race20080503_24h.htm[/url]

mdettweiler 2008-05-05 06:04

[quote=gd_barnes;132670]Xyzzy is being his funny self again...my avitar is upside down. lmao[/quote]
Now it's laying on its side. :wink:

Edit: And now, sometimes it's on its side and sometimes it's upside down, and sometimes it's right-side up. I think it's rotating all 4 directions now! lmao

gd_barnes 2008-05-05 16:19

[quote=kar_bon;132745]the scoring and some data about the third NPLB-race can be found here:
[URL]http://www.rieselprime.org/nplb/NPLB_Race20080503_24h.htm[/URL][/quote]

Karsten,

Thanks for compiling that. One thing: You might check the times on the graph. The rally ran from 19:00 GMT May 3rd thru 19:00 GMT May 4th. You are showing 15:00 GMT - 15:00 GMT. 15:00 is the time in Eastern Daylight time (EDT) U.S. that the rally ran.

I think you either need to change it to show 15:00 EDT-15:00 EDT, change it to 19:00 GMT-19:00 GMT, or change the graph bars because it's possibly reflecting testing from the incorrect hours.


Thanks,
Gary

gd_barnes 2008-05-05 16:57

Many thanks to everyone for participating in the rally! It was an unqualified success!! I'd like to shoot for a rally at least once/month from this point forward. We had held off on this one to allow the admins to catch their breath on processing, verifying, reporting, and updating results files, posts, and web pages after we had several rallies to get the project moving quickly at its outset.

Also, many thanks to Adam's (AES's) server. It handled a very large load for at least 24 hours without a glitch.

For as many k/n pairs as we tested, I would have expected us to find ~5 primes vs. the 3 that were actually found but obviously that is a normal fluctuation in the universe of the almost random distribution of primes.

We brought our testing up to n=423.7K for drive 1 with the rally and with the ongoing testing since the end of it, we're already at n=424.6K as of this moment.

We are shooting for n=600K on both drives 1 and 3 by year end so everyone keep plugging away and we'll make it! :smile:


Gary

kar_bon 2008-05-05 19:56

[QUOTE=gd_barnes;132777]
One thing: You might check the times on the graph. The rally ran from 19:00 GMT May 3rd thru 19:00 GMT May 4th. You are showing 15:00 GMT - 15:00 GMT. 15:00 is the time in Eastern Daylight time (EDT) U.S. that the rally ran.
[/QUOTE]

Done.


All times are UTC. The time now is 02:48.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.