How BOINC distributes work units

Message boards : BOINC client : How BOINC distributes work units
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile mitrichr
Avatar

Send message
Joined: 21 May 07
Posts: 349
United States
Message 19766 - Posted: 26 Aug 2008, 18:41:05 UTC

I am running four computers, and a lot of the same projects across a number of them. I run the screen savers, so I am able to see pretty much what is happening at any one time when I come into the rooms where the computers are. Often, when I come in and look, I will see that several computers are running WU's from the same project. Often, I will not see any WU's from one project or another, for quite some time. And, I am aware that BOINC operates based upon the weighting I have configured for each computer. I am also aware from the RSS feeds of news that some projects will halt new WU's for some reason for a time. So, my question is, when I see several computers running Rice (from WCG) or Spinhenge, or whatever, is this just random, or can I conclude that the project has put up a bunch of WU's at a specific time and that BOINC is at that time distributing the WU's based upon the project having put them up?

BTW, nice new look to the forum.

>>RSM
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

ID: 19766 · Report as offensive
Profile mitrichr
Avatar

Send message
Joined: 21 May 07
Posts: 349
United States
Message 19777 - Posted: 27 Aug 2008, 1:17:01 UTC - in response to Message 19776.  

So, my question is, when I see several computers running Rice (from WCG) or Spinhenge, or whatever, is this just random, or can I conclude that the project has put up a bunch of WU's at a specific time and that BOINC is at that time distributing the WU's based upon the project having put them up?

BTW, nice new look to the forum.

>>RSM


If you've been requesting work from a project but haven't been getting any work from that project for an extended period (for example they may be offline) then that project will accumulate a relatively large credit. Credit is the sum of the debts owed to it by your other projects. When that project comes back online and starts sending work again then yes, you would likely see all of your computers crunching that project for a while to collect the debt owed to that project.

If your machines have been getting work whenever they request it then all of your machines crunching the same project at the same time would be a coincidence. I doubt you would see that very often unless, of course, that project is given a high resource share on all your hosts.


I understand the weighting (Sek), I think Dag is getting it right for me, I did not think about projects which are not putting up WU's accumulating credit. I do see projects telling us in the RSS feeds that there will be nothing for some time, a few days, whatever, so that is probably the origin.

Thanks for the quick replies.

>>RSM
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

ID: 19777 · Report as offensive
Profile mitrichr
Avatar

Send message
Joined: 21 May 07
Posts: 349
United States
Message 19785 - Posted: 27 Aug 2008, 4:36:29 UTC - in response to Message 19780.  

I understand the weighting (Sek), I think Dag is getting it right for me, I did not think about projects which are not putting up WU's accumulating credit. I do see projects telling us in the RSS feeds that there will be nothing for some time, a few days, whatever, so that is probably the origin.

Thanks for the quick replies.

>>RSM

If you understand weighting and credit, then I suppose you would understand vouchers too. The vouchers for projects are there and held out to projects that have not used them "come and get them, when you got work". Each machine is a unique entity unaware of the other. Never mind



I do understand that credit for any project on any machine is irrespective of credit for the same project any other machine. I do not understand "never mind".

>>RSM
http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

ID: 19785 · Report as offensive
Pepo
Avatar

Send message
Joined: 3 Apr 06
Posts: 547
Slovakia
Message 19931 - Posted: 3 Sep 2008, 12:28:26 UTC - in response to Message 19785.  

Never mind

I do not understand "never mind".

>>RSM

Never mind then ;-)

when I see several computers running Rice (from WCG) or Spinhenge, or whatever, is this just random, or can I conclude that the project has put up a bunch of WU's at a specific time

I think that even if the project servers would manage to keep a steady flow of WUs of any type, BOINC client would still download them in bunches (because of the low-high watermark behavior of Long Term Debts) and possibly interleave them later during the computation (according to Short Term Debts).

Peter
ID: 19931 · Report as offensive
Profile mitrichr
Avatar

Send message
Joined: 21 May 07
Posts: 349
United States
Message 19944 - Posted: 3 Sep 2008, 20:19:55 UTC - in response to Message 19931.  

Never mind

I do not understand "never mind".

>>RSM

Never mind then ;-)

when I see several computers running Rice (from WCG) or Spinhenge, or whatever, is this just random, or can I conclude that the project has put up a bunch of WU's at a specific time

I think that even if the project servers would manage to keep a steady flow of WUs of any type, BOINC client would still download them in bunches (because of the low-high watermark behavior of Long Term Debts) and possibly interleave them later during the computation (according to Short Term Debts).

Peter


Thanks. What I had missed was that all of the machines can be building up credit at the same time, and so the fulfillment of that credit can occur at the same time.

>>RSM

http://sciencesprings.wordpress.com
http://facebook.com/sciencesprings

ID: 19944 · Report as offensive

Message boards : BOINC client : How BOINC distributes work units

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.