NPLB LLRnet server discussion - Page 18

henryzz · 2008-03-04, 07:27

Quote:

Originally Posted by IronBits

I have as many ports as you like, and I agree with Carlos...
There are not very many crunchers on this project to pass them around to several servers.
In about a week, I will have a full time dedicated commercial circuit installed with 10M down / 1M up with unlimited data usage.
If necessary, I can change that to 15M down / 2M up. :)
Once that gets into place, I can host as many server ports as you see fit to run.
The windows box running port 500 will get moved over to the new CentOS Server I just built (quad/8gb ram/over 1/2TB storage) after the circuit gets installed.
I'll be putting up web pages and stats for each server/port so you will always know how many knpairs each port has, the results.txt files will get moved to a password protected directory so you can fetch them any time you like.
I will also set up a mechanism so that you can put work in a passworded directory, and it will take that server/port down, append the data to knpairs.txt then bring the server/port back online.
I also read there is a way being developed to auto-notify users when a prime is found. Once that gets worked out, it should be pretty much automated. :)
Let me know if you can think of anything else you want or can think of, that would make your lives easier.
Keep on crunching!

sounds good

gd_barnes · 2008-03-04, 07:39

Quote:

Originally Posted by IronBits

I have as many ports as you like, and I agree with Carlos...
There are not very many crunchers on this project to pass them around to several servers.
In about a week, I will have a full time dedicated commercial circuit installed with 10M down / 1M up with unlimited data usage.
If necessary, I can change that to 15M down / 2M up. :)
Once that gets into place, I can host as many server ports as you see fit to run.
The windows box running port 500 will get moved over to the new CentOS Server I just built (quad/8gb ram/over 1/2TB storage) after the circuit gets installed.
I'll be putting up web pages and stats for each server/port so you will always know how many knpairs each port has, the results.txt files will get moved to a password protected directory so you can fetch them any time you like.
I will also set up a mechanism so that you can put work in a passworded directory, and it will take that server/port down, append the data to knpairs.txt then bring the server/port back online.
I also read there is a way being developed to auto-notify users when a prime is found. Once that gets worked out, it should be pretty much automated. :)
Let me know if you can think of anything else you want or can think of, that would make your lives easier.
Keep on crunching!

Very nice IronBits! I and I'm sure everyone appreciates this kind of dedication to helping things run smoothly!

Thanks!

Gary

AMDave · 2008-03-04, 09:06

Speaking of that commercial circuit, I'd like to refer to the problem experienced in the last rally.

I ran some calculations and I suspect that the number of hits per hour (I estimated over 2100/hr) on CRUS could have triggered a either a NAT throttle by the ISP monitor or a DDoS prevention routine.

The NAT throttle is not something they would readily tell you about, but it could explain why it appeared to choke, as might a DDoS prevention routine.

I have witnessed similar occurences that were resolved to these triggers.

Even though RAM was a little tight I would not have expected it to get locked up the way it appeared to, especially after the connection limit was increased, given that the machine is quite capable.

I envision a scenario where the connections opened at a rate higher than a NAT/DDoS throttle would allow the machine to deal with the associated I/O.

Just a hypothesis.

em99010pepe · 2008-03-11, 12:13

I have two work machines that aren't mine working on my server port 100 and soon those machines will be away from my office. We are moving from offices and I'll be separated from my friends with less time to control their machines. I suppose the server port 100 will be shutdown after it dries but port 300 will be up for awhile, correct? Just wanted to be sure because I need urgently to change both ports from 100 to 300.

Carlos

AMDave · 2008-03-11, 12:53

I moved mine off 100 yesterday.

I think they are gary's per his post here
http://www.mersenneforum.org/showthr...382#post128382

mdettweiler · 2008-03-11, 16:02

Quote:

Originally Posted by em99010pepe

I have two work machines that aren't mine working on my server port 100 and soon those machines will be away from my office. We are moving from offices and I'll be separated from my friends with less time to control their machines. I suppose the server port 100 will be shutdown after it dries but port 300 will be up for awhile, correct? Just wanted to be sure because I need urgently to change both ports from 100 to 300.

Carlos

If I remember correctly, you wanted to dry out both 100 and 300 and then shut them down. Probably your best bet, if you want a long-term solution, is IronBits' port 5000 server, as that one will probably be up for a LOOOOONG time.

em99010pepe · 2008-03-11, 22:29

16 000 k/n pairs left on my server port 100. Need help to dry it. I suppose we only need 3 days to do it.

mdettweiler · 2008-03-11, 22:36

Quote:

Originally Posted by em99010pepe

16 000 k/n pairs left on my server port 100. Need help to dry it. I suppose we only need 3 days to do it.

I've currently got a Pentium3 550Mhz working on it...yes, I know, not much at all.

Maybe later today I'll be able to put one core of my Core 2 Duo on it.

em99010pepe · 2008-03-11, 22:49

I have 2 cores on it, other 6 cores are finishing some lower ranges from drive 3, just need one more day then you guys can move the cores to another server while I dry by myself the server.

IronBits · 2008-03-11, 22:52

I just put 7 cores on it, but Boinc is running as well.

mdettweiler · 2008-03-12, 00:32

I've now got one core of my Core 2 Duo on port 100 LLRnet.

More precisely, I used LLRnet to grab 500 k/n pairs, and I am now crunching them with LLR 3.7.1c (when they're done I'll run my script on the lresults file to generate an LLRnet-compatible tosend file).

A little less convenient than the normal method, yes, but I'll happily sacrifice a little convenience for a nice sized speed boost.

2008-03-11, 22:29	#194
em99010pepe Sep 2004 2·5·283 Posts	16 000 k/n pairs left on my server port 100. Need help to dry it. I suppose we only need 3 days to do it. Last fiddled with by em99010pepe on 2008-03-11 at 22:32

2008-03-11, 22:49	#196
em99010pepe Sep 2004 2×5×283 Posts	I have 2 cores on it, other 6 cores are finishing some lower ranges from drive 3, just need one more day then you guys can move the cores to another server while I dry by myself the server. Last fiddled with by em99010pepe on 2008-03-11 at 22:50

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
LLRnet servers for NPLB	kar_bon	No Prime Left Behind	1343	2014-08-20 09:38
NPLB LLRnet server #1 - dried	em99010pepe	No Prime Left Behind	19	2008-03-26 06:19
New LLRnet servers discussion	IronBits	Conjectures 'R Us	11	2008-03-20 03:43
CRUS LLRnet server discussion	em99010pepe	Conjectures 'R Us	181	2008-02-04 19:51
llrnet server down?	Molnar	Prime Sierpinski Project	4	2006-10-19 14:32

2008-03-04, 09:06	#190
AMDave Jan 2006 deep in a while-loop 292₁₆ Posts	Speaking of that commercial circuit, I'd like to refer to the problem experienced in the last rally. I ran some calculations and I suspect that the number of hits per hour (I estimated over 2100/hr) on CRUS could have triggered a either a NAT throttle by the ISP monitor or a DDoS prevention routine. The NAT throttle is not something they would readily tell you about, but it could explain why it appeared to choke, as might a DDoS prevention routine. I have witnessed similar occurences that were resolved to these triggers. Even though RAM was a little tight I would not have expected it to get locked up the way it appeared to, especially after the connection limit was increased, given that the machine is quite capable. I envision a scenario where the connections opened at a rate higher than a NAT/DDoS throttle would allow the machine to deal with the associated I/O. Just a hypothesis.

2008-03-11, 12:13	#191
em99010pepe Sep 2004 B0E₁₆ Posts	I have two work machines that aren't mine working on my server port 100 and soon those machines will be away from my office. We are moving from offices and I'll be separated from my friends with less time to control their machines. I suppose the server port 100 will be shutdown after it dries but port 300 will be up for awhile, correct? Just wanted to be sure because I need urgently to change both ports from 100 to 300. Carlos

2008-03-11, 12:53	#192
AMDave Jan 2006 deep in a while-loop 1222₈ Posts	I moved mine off 100 yesterday. I think they are gary's per his post here http://www.mersenneforum.org/showthr...382#post128382

2008-03-11, 22:52	#197
IronBits I ♥ BOINC! Oct 2002 Glendale, AZ. (USA) 2131₈ Posts	I just put 7 cores on it, but Boinc is running as well.

2008-03-12, 00:32	#198
mdettweiler A Sunny Moo Aug 2007 USA (GMT-5) 3×2,083 Posts	I've now got one core of my Core 2 Duo on port 100 LLRnet. More precisely, I used LLRnet to grab 500 k/n pairs, and I am now crunching them with LLR 3.7.1c (when they're done I'll run my script on the lresults file to generate an LLRnet-compatible tosend file). A little less convenient than the normal method, yes, but I'll happily sacrifice a little convenience for a nice sized speed boost.