Welcome to MilkyWay@home

Project down for maintenance??

Message boards : Number crunching : Project down for maintenance??
Message board moderation

To post messages, you must log in.

AuthorMessage
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 302,524,931
RAC: 15
Message 5438 - Posted: 10 Oct 2008, 21:08:44 UTC

10/10/2008 2:01:36 PM|Milkyway@home|Sending scheduler request: Requested by user. Requesting 0 seconds of work, reporting 1 completed tasks
10/10/2008 2:01:41 PM|Milkyway@home|Scheduler request succeeded: got 0 new tasks
10/10/2008 2:01:41 PM|Milkyway@home|Message from server: Project is temporarily shut down for maintenance

??
ID: 5438 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dave Przybylo
Avatar

Send message
Joined: 5 Feb 08
Posts: 236
Credit: 49,648
RAC: 0
Message 5439 - Posted: 10 Oct 2008, 21:14:37 UTC

hopefully you've noticed we finally upgraded to the newest version of BOINC!
Dave Przybylo
MilkyWay@home Developer
Department of Computer Science
Rensselaer Polytechnic Institute
ID: 5439 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 302,524,931
RAC: 15
Message 5440 - Posted: 10 Oct 2008, 21:36:09 UTC - in response to Message 5439.  

Indeed -- you look like all the other guys now <smile>.

By the way, regarding the fun and games on credits (I am NOT using an optimized application -- just have a lot of spare horsepower out there), a concern I have is that the large 'credit per second' awards will get you folks noticed by the much larger BOINC community and that might catch you some flack here.

What I'm talking about is that even with the unoptimized application generally available here -- which I use -- I get something like 40 to 60 credits an hour per core (AMD 4200 to AMD 9850).

On the SETI project, even with the optimized application (which *anyone* can download) -- I get 25 to 30 credits/hour. On Einstein (which doesn't have an optimized application), the number is about half that -- and that is pretty much the same for other projects (Spinhenge, Rosetta, Malaria) as well. Climate seems a bit better (20 to 25 with an unoptimized client).

Now with the hand coded personally optimized clients that a very few have figured out for Milkyway, we're talking what appears to be maybe 20 to 50 times that amount (>1K credits per hour per core). That sort of number is definitely going to wake up the ghosts of BOINC past, current and future and may well result in a level of attention you might not desire.




ID: 5440 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dave Przybylo
Avatar

Send message
Joined: 5 Feb 08
Posts: 236
Credit: 49,648
RAC: 0
Message 5441 - Posted: 10 Oct 2008, 21:46:45 UTC

Indeed we have a conundrum. If we improve the code, it will be faster and we will have to lower credits which people dont like. If we don't improve our code, we get yelled at for wasting cpu time.
Dave Przybylo
MilkyWay@home Developer
Department of Computer Science
Rensselaer Polytechnic Institute
ID: 5441 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Milksop at try

Send message
Joined: 1 Oct 08
Posts: 106
Credit: 24,162,445
RAC: 0
Message 5442 - Posted: 10 Oct 2008, 21:51:26 UTC - in response to Message 5441.  
Last modified: 10 Oct 2008, 21:53:36 UTC

That's easy. The people are concerned by the credits/hour they get, not the credits/WU. If the code is 50 times as fast, the credits can also be lowered by a factor of 50.
If you want, I could do some flops counting ;)
ID: 5442 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Dave Przybylo
Avatar

Send message
Joined: 5 Feb 08
Posts: 236
Credit: 49,648
RAC: 0
Message 5443 - Posted: 10 Oct 2008, 22:52:18 UTC - in response to Message 5442.  

Well we're improving the code. We did in fact implement a few of the changes you pointed out. If you'd like to point out more we'd be happy to take a look and implement those if they pan out as well. But we are looking into our code.
Dave Przybylo
MilkyWay@home Developer
Department of Computer Science
Rensselaer Polytechnic Institute
ID: 5443 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
BarryAZ

Send message
Joined: 1 Sep 08
Posts: 520
Credit: 302,524,931
RAC: 15
Message 5444 - Posted: 10 Oct 2008, 22:54:06 UTC - in response to Message 5442.  

Actually, the credits per hour is the concern for some folks -- so what might well happen is if the released code is 50 times faster, the credit per work unit would be reduced by that factor -- no net difference to users with public code. Folks with private enhanced code would see the reduction though. Shouldn't cause them concern though since they are not worried about their credit/hour rates and are simply looking to make sure that the most efficient code is available.

ID: 5444 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Milksop at try

Send message
Joined: 1 Oct 08
Posts: 106
Credit: 24,162,445
RAC: 0
Message 5445 - Posted: 10 Oct 2008, 23:27:31 UTC - in response to Message 5443.  

Well we're improving the code. We did in fact implement a few of the changes you pointed out. If you'd like to point out more we'd be happy to take a look and implement those if they pan out as well. But we are looking into our code.

That's good to hear. Maybe I won't have time this weekend (I guess that's ok for you ;), but next week there will be more suggestions.

In the meantime we detached 5 quads. Now only some duals are holding position here. I don't know if you like it or not (the quads had quite some throughput), but maybe it's a small relief for some others. If some substantial improvement arrives at the official app, we will stop crunching for this account as promised in the profile.
ID: 5445 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
EigenState

Send message
Joined: 19 Apr 08
Posts: 13
Credit: 151,588
RAC: 0
Message 5446 - Posted: 10 Oct 2008, 23:59:20 UTC - in response to Message 5441.  

Greetings,

Indeed we have a conundrum. If we improve the code, it will be faster and we will have to lower credits which people dont like. If we don't improve our code, we get yelled at for wasting cpu time.


A conundrum? I beg to differ!

Credits are by definition an artificial measure of a participant's contribution to the project that can be manipulated with great flexibility. The only valid concern is the scientific output of the project per unit time. It is self-evident that an optimized code offers the potential to greatly increase that scientific output and thereby greatly enhance the potential of this particular research endeavor to become a meaningful contribution to physics.

Greater productivity (research results) can be fully anticipated to lead to more papers being published and more talks at conferences. That is quite clearly in the best interests of the members of the research team.

You have several participants that appear to have successfully optimized your code. It behooves you to do everything possible to work with them to develop an optimized code that can and will be made available to all of the volunteer participants. The failure to do so is an implicit statement that you really do not care about the physics, nor about the resources that your volunteers bring to bear on this project.

I can only imagine my response were one of my research students to inform me that they had taken orders of magnitude more time than necessary to conduct an experiment!

Best regards,
EigenState






ID: 5446 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
John Clark

Send message
Joined: 4 Oct 08
Posts: 1734
Credit: 64,228,409
RAC: 0
Message 5449 - Posted: 11 Oct 2008, 8:45:18 UTC
Last modified: 11 Oct 2008, 8:46:02 UTC

EigenState made a good point in favour of the co-operation now taking place with the code efficiency improvers (optimisers).

I hope the endevour progresses and I look forwards, in due course, to the announcement, and link, to the faster code when it comes. Clearly the credit granted can, and should, be reduced appropriately against the estimate of the average crunching machine believed to exist in the project.
ID: 5449 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Logan
Avatar

Send message
Joined: 15 Aug 08
Posts: 163
Credit: 3,876,869
RAC: 0
Message 5450 - Posted: 11 Oct 2008, 9:58:07 UTC
Last modified: 11 Oct 2008, 10:31:54 UTC

I would remember you what Crunch3r was done a lot of wu's (a lot of correct science, not like stock app) and only awarded with 10% of the credits by the (absurd) credits limit (as if have done 10% of science). That is the reward to do a lot of science work in this project...?

I think that it would be fair if it only awarded by 10% from the work, you just get the 10% of it... and don't take profit of the 90% remaining...

Best regards.
Logan.

BOINC FAQ Service (Ahora, también disponible en Español/Now available in Spanish)
ID: 5450 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : Project down for maintenance??

©2024 Astroinformatics Group