Switched to Einstine


log in

Advanced search

Message boards : Number crunching : Switched to Einstine

1 · 2 · Next
Author Message
Profile Jim Bohan
Avatar
Send message
Joined: 23 Dec 01
Posts: 47
Credit: 19,510,705
RAC: 1,843
United States
Message 1307638 - Posted: 19 Nov 2012, 7:07:05 UTC

Ok, been putting up with this B---S---t for too long, I have been with SETI for years and although I understand that computers and programming have issues at times my computers are not gettng any work and I have to do the NNT all the time to send work (only GPU units, nothing else)
I've gone to Einstine for now. When you guys get this problem figured out let me know and I'll come back.

<<Jim >>
____________
Member
B-52 Stratofortress
Association
Retired Air Force

Profile Gary CharpentierProject donor
Volunteer tester
Avatar
Send message
Joined: 25 Dec 00
Posts: 13207
Credit: 7,985,876
RAC: 16,130
United States
Message 1307724 - Posted: 19 Nov 2012, 14:55:12 UTC - in response to Message 1307638.

Ok, been putting up with this B---S---t for too long, I have been with SETI for years and although I understand that computers and programming have issues at times my computers are not gettng any work and I have to do the NNT all the time to send work (only GPU units, nothing else)
I've gone to Einstine for now. When you guys get this problem figured out let me know and I'll come back.

<<Jim >>

Thank you. Your switch to Einstein has help alleviate the problem. In case you didn't know the problem is to many crunchers with too many fast crunch cards. Every user that switches makes the problem a little less.

____________

Cherokee150
Send message
Joined: 11 Nov 99
Posts: 112
Credit: 25,682,748
RAC: 7,369
United States
Message 1307727 - Posted: 19 Nov 2012, 15:14:59 UTC - in response to Message 1307638.
Last modified: 19 Nov 2012, 15:22:17 UTC

Hi Jim,
You asked us to let you know when they have figured things out. They have, and have even determined a fix. It will take awhile to implement, possibly a few weeks to a month or more, but they are working on it. They were in the lab even late last night (Sunday) when they reported to us in detail (http://setiathome.berkeley.edu/forum_thread.php?id=70080).

To put it in a nutshell, Jim, so many people upgraded their systems and added GPUs so fast that the system was overloaded with requests, kind of like a DOS attack! I guess the SETI staff now knows what it's like to be a rock star, when a thousand fans rush to get autographs all at once (ask Matt when he gets back from his world tour)! lol ;)

One of the wonderful features of BOINC, Jim, is you can leave SETI as-is in BOINC, add Einstein, and let BOINC determine which application has available units. When SETI comes back up fixed, BOINC will start feeding you units again. At that time you can either switch Einstein to "No new tasks", or leave it on if you wish to continue processing both.

Thank you so much for all you contributions to SETI, Jim! I, and I am sure the SETI staff, truly appreciate your long-time contribution to science.

Profile S@NL - eFMer - efmer.com/boincProject donor
Volunteer tester
Avatar
Send message
Joined: 7 Jun 99
Posts: 512
Credit: 131,446,681
RAC: 28,879
United States
Message 1307735 - Posted: 19 Nov 2012, 16:02:04 UTC - in response to Message 1307733.

I can't say I am happy with the current state of things here on my favorite project, but I do know that they shall get better. After all, with over 13 years under my belt here, I have seen a lot of ups and downs.

Helps is on its way, the simple trick. Make the workunits a lot longer to process, by a factor of 4 or so.
This will give us a year or so.......
____________
TThrottle Control your temperatures. BoincTasks The best way to view BOINC. Anza Borrego Desert hiking.

spitfire_mk_2
Avatar
Send message
Joined: 14 Apr 00
Posts: 463
Credit: 13,238,246
RAC: 7,687
United States
Message 1307738 - Posted: 19 Nov 2012, 16:10:01 UTC - in response to Message 1307638.

Ok, been putting up with this B---S---t for too long, I have been with SETI for years and although I understand that computers and programming have issues at times my computers are not gettng any work and I have to do the NNT all the time to send work (only GPU units, nothing else)
I've gone to Einstine for now. When you guys get this problem figured out let me know and I'll come back.

<<Jim >>

Same here. I run both. Since SETI is down, again, Einstein is running.
____________

juan BFBProject donor
Volunteer tester
Avatar
Send message
Joined: 16 Mar 07
Posts: 5489
Credit: 316,554,304
RAC: 132,316
Brazil
Message 1307742 - Posted: 19 Nov 2012, 16:17:40 UTC - in response to Message 1307724.
Last modified: 19 Nov 2012, 16:41:49 UTC

Thank you. Your switch to Einstein has help alleviate the problem. In case you didn't know the problem is to many crunchers with too many fast crunch cards. Every user that switches makes the problem a little less.

You could count with my help with 7 hungry hosts. :(

To avoid the dissease they are killing the patient... something cleary out off maind!

discussed once or twice in the past, and it was noted that we were starting to approach the limits of mysql.


If that was a "time bomb" well know problem, why dont´t make nothing to avoid before it´s exploding?

Matt please, please... I beg you, return as soon as you can...
____________

WezH
Volunteer tester
Send message
Joined: 19 Aug 99
Posts: 252
Credit: 6,146,276
RAC: 46,010
Finland
Message 1307743 - Posted: 19 Nov 2012, 16:19:05 UTC
Last modified: 19 Nov 2012, 16:19:57 UTC

It will be interesting to see how would Einstein handle load IF heavyweight crunchers like msattler and Juan decides to switch to Einstein with full capacity until Seti is back up.
____________
"Please keep Your signature under four lines so Internet traffic doesn't go up too much"

- In 1992 when I had my first e-mail address -

juan BFBProject donor
Volunteer tester
Avatar
Send message
Joined: 16 Mar 07
Posts: 5489
Credit: 316,554,304
RAC: 132,316
Brazil
Message 1307749 - Posted: 19 Nov 2012, 16:26:22 UTC - in response to Message 1307743.
Last modified: 19 Nov 2012, 16:52:34 UTC

It will be interesting to see how would Einstein handle load IF heavyweight crunchers like msattler and Juan decides to switch to Einstein with full capacity until Seti is back up.


As you could notice, I all ready switch to Einstein (0 resource for now) and i know Mark and Tbret do the same, we 3 alone have more than 1.2 MM in SETI RAC, so by doing that we will help to fix the SETI problem... Firehawk allready switch to GPUGrid, TPL to collantz, Hope Russ do something like that soon (don´t have any ideia to what project for now)... Janice/Steve ithink going to Einstein to, anyone else not shure... At least a combinned force of more than 2.5MM in SETI RAC.

But be clear to all, i´m not happy to do that. The way the things takes, everyone will think we are the source of the problem, exactly the oposite, we just trying to follow a dream (Help to find an ET) and do our best for the project... nobody tell us "you can´t put to many crunchers to work or the project will be down..."

(edit)

Follow the kitties, my friend. Follow the kitties.


I all ready do that, allways follow the kitties path... You know, I´m a kitties fan! :)

I´m going to put 2 new 690 to crunch today... so count with some more help from my side... Hope Einstein could welcome us without cause any problem to them, thats not my ideia, just want to contribute with science and Try to help to find a ET... snif!
____________

Profile Bernie Vine
Volunteer moderator
Volunteer tester
Avatar
Send message
Joined: 26 May 99
Posts: 7199
Credit: 29,231,082
RAC: 36,097
United Kingdom
Message 1307752 - Posted: 19 Nov 2012, 16:43:52 UTC

A little factlet for consideration

The top 20 machines of SETI@Home have between them

172 CPU's, mostly i7's and i5's

89 GPU's most probably high end NVIDIA

With a combined RAC of over 2 million

That's just 20 machines. Makes you think!!
____________


Today is life, the only life we're sure of. Make the most of today.

Profile Tron
Send message
Joined: 16 Aug 09
Posts: 180
Credit: 2,236,055
RAC: 0
United States
Message 1307824 - Posted: 19 Nov 2012, 18:54:03 UTC

172 CPU's, mostly i7's and i5's
89 GPU's most probably high end NVIDIA
With a combined RAC of over 2 million
That's just 20 machines. Makes you think!!


Makes Me think : I have alot of work to do if I plan on surpassing them. mwuahahaha!

Brkovip
Avatar
Send message
Joined: 18 May 99
Posts: 274
Credit: 144,370,125
RAC: 493
United States
Message 1307835 - Posted: 19 Nov 2012, 19:07:26 UTC - in response to Message 1307824.

172 CPU's, mostly i7's and i5's
89 GPU's most probably high end NVIDIA
With a combined RAC of over 2 million
That's just 20 machines. Makes you think!!


Makes Me think : I have alot of work to do if I plan on surpassing them. mwuahahaha!


Me thinks you better be making a lot of money. The electricity bill alone with that many computers would kill most.
____________

mikeej42
Send message
Joined: 26 Oct 00
Posts: 109
Credit: 790,756,468
RAC: 230
United States
Message 1307842 - Posted: 19 Nov 2012, 19:37:32 UTC - in response to Message 1307747.

I'll reiterate my previous advice.
Set Einstein to zero workshare, and leave Seti as is. Your computers will automagically return to Seti when work is available.
Follow the kitties, my friend. Follow the kitties.


Einstein tends to eat up a lot more disk space and setting Einstein to 0 workshare will control the number of work units requested. I had hundreds of systems almost run out of diskspace the last time SETI had a prolonged outage and all my machines started trying to keep 10 day caches of Einstein work units.

Follow the kitties advice and all will be right with the world....
____________

Profile HAL9000
Volunteer tester
Avatar
Send message
Joined: 11 Sep 99
Posts: 4665
Credit: 123,818,858
RAC: 94,900
United States
Message 1307844 - Posted: 19 Nov 2012, 19:43:31 UTC - in response to Message 1307824.

172 CPU's, mostly i7's and i5's
89 GPU's most probably high end NVIDIA
With a combined RAC of over 2 million
That's just 20 machines. Makes you think!!


Makes Me think : I have alot of work to do if I plan on surpassing them. mwuahahaha!

If you look at the top 20 participants the combined RAC is a lot higher. Totaling about 7,353,253. However that is only 4.7% of the total SETI@Home RAC.

BOINCstats shows 268,760 active users for BOINC combined & 151,581 active users for SETI@Home. Einstein by comparison only has 36,829 active users.

I can imagine that if any of the other projects has a sudden influx of 100,000 users their systems would show similar strain.
____________
SETI@home classic workunits: 93,865 CPU time: 863,447 hours

Join the BP6/VP6 User Group today!

Richard HaselgroveProject donor
Volunteer tester
Send message
Joined: 4 Jul 99
Posts: 8833
Credit: 53,666,093
RAC: 48,650
United Kingdom
Message 1307849 - Posted: 19 Nov 2012, 19:51:05 UTC - in response to Message 1307842.

I'll reiterate my previous advice.
Set Einstein to zero workshare, and leave Seti as is. Your computers will automagically return to Seti when work is available.
Follow the kitties, my friend. Follow the kitties.

Einstein tends to eat up a lot more disk space and setting Einstein to 0 workshare will control the number of work units requested. I had hundreds of systems almost run out of diskspace the last time SETI had a prolonged outage and all my machines started trying to keep 10 day caches of Einstein work units.

Follow the kitties advice and all will be right with the world....

It's their CPU application (S6LV1) which tends to eat up disk space: there are enough other CPU projects out there that we can probably spread the load more widely if this current outage turns out to be prolonged. S6LV1 is also approaching a transition point (45 days work remaining, less if we all rush over there and gobble it up) - things tend to get a bit spotty there with large downloads as they fill in the final gaps.

I suspect that more people here will go to Einstein to run their CUDA apps. They don't hog disk space the way their CPU apps do, but you need a robust ISP and a generous data allowance - my GTX 670 is downloading about 60MB an hour (two tasks at a time, about 30 minutes runtime) when crunching Einstein. And that can lead to a significant gap between jobs (as one set of files uploads, and the next set downloads) when running Resource Share zero.

juan BFBProject donor
Volunteer tester
Avatar
Send message
Joined: 16 Mar 07
Posts: 5489
Credit: 316,554,304
RAC: 132,316
Brazil
Message 1307859 - Posted: 19 Nov 2012, 20:20:02 UTC - in response to Message 1307849.
Last modified: 19 Nov 2012, 20:27:34 UTC

As allways, Richard is right.

I just put 0 resources to start all the hosts because i belive if i put them all together could be a problem to get all the WU they need. I don´t want to be the origin in any trouble with Einstein, as people think we are the source of the SETI problem.

Will increase slowly the buffer on my hosts, one at a time to avoid any issue. On other hand even if i have diferent 3 ISP with 10MBps from each, is wise to avoid any problem with the DL because i know the Einstein WU are very big.

Regards the disk space, don´t worry i have a lot of Terabytes space avaiable in our servers, diferent from the lab, here i never allow our systems using more than 50% of the avaiable resources at anytime, exactly because that we could run on a 24/7 cicle, without any interruption, or i loose my job...
____________

mikeej42
Send message
Joined: 26 Oct 00
Posts: 109
Credit: 790,756,468
RAC: 230
United States
Message 1307862 - Posted: 19 Nov 2012, 20:32:52 UTC - in response to Message 1307849.

It's their CPU application (S6LV1) which tends to eat up disk space: there are enough other CPU projects out there that we can probably spread the load more widely if this current outage turns out to be prolonged. S6LV1 is also approaching a transition point (45 days work remaining, less if we all rush over there and gobble it up) - things tend to get a bit spotty there with large downloads as they fill in the final gaps.

I suspect that more people here will go to Einstein to run their CUDA apps. They don't hog disk space the way their CPU apps do, but you need a robust ISP and a generous data allowance - my GTX 670 is downloading about 60MB an hour (two tasks at a time, about 30 minutes runtime) when crunching Einstein. And that can lead to a significant gap between jobs (as one set of files uploads, and the next set downloads) when running Resource Share zero.


Yes I only have 1 system that has a CUDA card in it. 99.9999... percent of the work I have done was CPU only. I have resigned myself to crunch on what the large GPU systems leave behind. I do what I can with what I have access to.

Still not a bad RAC considering.
____________

Profile Link
Avatar
Send message
Joined: 18 Sep 03
Posts: 841
Credit: 1,578,326
RAC: 52
Germany
Message 1307888 - Posted: 19 Nov 2012, 21:42:18 UTC - in response to Message 1307849.
Last modified: 19 Nov 2012, 21:48:02 UTC

my GTX 670 is downloading about 60MB an hour (two tasks at a time, about 30 minutes runtime) when crunching Einstein. And that can lead to a significant gap between jobs (as one set of files uploads, and the next set downloads) when running Resource Share zero.

You can trick BOINC to cache an additional WU if you set <count> to 0.99 (if running 1 WU at a time) and probably 0.49 if running 2 at a time. Probably because the second one I haven't tested, 0.99 works for me with Collatz as backup project for Milkyway perfectly.

I'm running 2 Milkyway WUs at once, when the first (of the last two) is finished, the first Collatz WU is cached, but not started, when the second MW WU is finished BOINC starts the Collatz WU and requests a second one since it thinks that the GPU is not completely in use. But once it has it, it won't start it, because that would be more than 1.00. Once the first Collatz WU ist finished, BOINC starts the next one, uploads the result and requests next WU. So I always have one waiting to start. With <count> 1.00 for Collatz I had always a break of about 15 seconds (every 20 minutes or so), because BOINC would first upload the result, than report and request new WU and finally download the new WU. That was not good enough for me.

Unfortunately this does not work for CPUs, BOINC would start both WUs.
____________
.

1 · 2 · Next

Message boards : Number crunching : Switched to Einstine

Copyright © 2014 University of California