| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
| World Community Grid Forums
|
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 5
|
|
| Author |
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
I've been crunching at WCG for over two and a half years now (since day 3 I think) and have not crunched on any other projects. My team had expanded its presence to some other DC projects and, as captain, it seemed to me that I should try to maintain something of a presence on those projects as well. Still, WCG is where I want to keep my focus and priority. I know how you can set different resource shares for each project but that only controls how your free CPU is "allocated" to the projects. Crunching on a WU is the "end" of the pipeline. I want to manage the "start" of the pipeline and limit the number of WUs I download from a project. In a sense, I want to "allocate" a share of my "queue" to each project. If I could use a different connect to server value for each project, I'd probably be fine but that appears to be a value I can only set in the default BOINC profile. at least on the other projects. If I use a two day connect to server setting, it looks like I get two days worth of work for EACH project and it looks like I am setting myself up for trouble that way. It appears that the default profile settings are common across projects but the settings in the home, work, school profiles are unique to each project but don't cover the connect to server value. Basically, if I'm limiting a project's resource share to 10 percent, I'd like to limit the WUs downloaded from that project to 10 percent of the total downloaded too. I'm sure I'm missing something pretty basic but that's what I get for crunching WCG only!
---------------------------------------- For those wondering, pulling a "Dolly Parton" refers to a rather petite and very buxom singer here who has described herself as 10 pounds of potatoes stuffed into a 5 pound sack. Thus, pulling a "Dolly Parton" refers to any type of attempt to feed more of something into a process than it can handle. |
||
|
|
retsof
Former Community Advisor USA Joined: Jul 31, 2005 Post Count: 6824 Status: Offline Project Badges:
|
If I use a two day connect to server setting, it looks like I get two days worth of work for EACH project and it looks like I am setting myself up for trouble that way. It seems to be a BOINC thing. The first time in, it doesn't realize that you are working on OTHER projects.After it connects the next time, it usually comes to its senses, and apportions workunits realistically after that. Another problem is that the estimates are wildly inaccurate on a few projects. It may say "50 hours" even though the workunit may complete in two. That keeps more workunits from being downloaded since that amount that can run is based on the due date. If I add another project, after the download starts, I change the project to "won't get new tasks" so more will not come down. I have even been known to abort extra workunits in the first set, again because it assumes that you only want that project. The proportion defaults to "100". I will usually go in and change that quickly to some other number if anything else is desired. I then join one of the teams (Vulture Central in my case, since there are over 50 now. I started most of them.) Hit "update project" and watch the proportion and team change on the BOINC advanced view. The new project can then be changed to "allow new tasks". Let's pretend SETI is the other project. The numbers are added to get a total for proportions. If WCG is 100 and SETI is 100, the total is "200". Each project gets 50% in that case. It is the same as if WCG were 50 and SETI were 50, or WCG 10 and SETI 10. Project mixing works best if you pick other project(s) with similar due dates. Some climate prediction projects are a matter of months, and some mathematical projects are a day.
SUPPORT ADVISOR
----------------------------------------Work+GPU i7 8700 12threads School i7 4770 8threads Default+GPU Ryzen 7 3700X 16threads Ryzen 7 3800X 16 threads Ryzen 9 3900X 24threads Home i7 3540M 4threads50% [Edit 3 times, last edit by retsof at Aug 29, 2007 12:26:33 PM] |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Were you talking about her latest maker-over.... scary?
----------------------------------------If you've set the share, which is a project specific value, and set up the client at all projects 'a' host is attached to, you need to consider that each time you change the home/work/school/default at any project, it will kick in for all projects. Now if you happen to do that change at WCG and hit the update button, obviously, it will pull at the skirts of WCG first. The client ignores the other projects momentarily as their scheduled contact is somewhere in the future. BOINC is keeping count of the amount of work in buffer, the crunching share and the amount of crunching it has done for each, plus the project switch frequency, which also takes account, from 5.8, with checkpoints. All in all whatever the mix is and however big Dolly has grown, after 14 days BOINC will have figured out who should get what and how much. Eventually if you happen to have a short deadline project and it has accumulated enough debt, BOINC will stop pulling work and deplete the time credit. Message is, just let it do its thing, really, and having CPDN and Tanpaku and a number with deadlines in between on a few machines, it does manage. If you attach, remove, change on a daily basis, BOINC will never figure out what you want! I think Didactylos has written it many times: Don't micro-manage.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
If I use a two day connect to server setting, it looks like I get two days worth of work for EACH project and it looks like I am setting myself up for trouble that way. It seems to be a BOINC thing. The first time in, it doesn't realize that you are working on OTHER projects.After it connects the next time, it usually comes to its senses, and apportions workunits realistically after that. Another problem is that the estimates are wildly inaccurate on a few projects. It may say "50 hours" even though the workunit may complete in two. That keeps more workunits from being downloaded since that amount that can run is based on the due date. If I add another project, after the download starts, I change the project to "won't get new tasks" so more will not come down. I have even been known to abort extra workunits in the first set, again because it assumes that you only want that project. The proportion defaults to "100". I will usually go in and change that quickly to some other number if anything else is desired. I then join one of the teams (Vulture Central in my case, since there are over 50 now. I started most of them.) Hit "update project" and watch the proportion and team change on the BOINC advanced view. The new project can then be changed to "allow new tasks". Let's pretend SETI is the other project. The numbers are added to get a total for proportions. If WCG is 100 and SETI is 100, the total is "200". Each project gets 50% in that case. It is the same as if WCG were 50 and SETI were 50, or WCG 10 and SETI 10. Project mixing works best if you pick other project(s) with similar due dates. Some climate prediction projects are a matter of months, and some mathematical projects are a day. Quite right, and thanks, Retsof. After the initial download when a project is first added, BOINC does seem to behave much better. As with so many other things about BOINC, giving it time is more often than not a good solution. Of course, heaven forbid that I abort or be late returning a WU so that initial flood evokes fears of crunching ineptitude from taking on more work than you can complete. Perhaps an addition to the WCG 12 step program. |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges:
|
Were you talking about her latest maker-over.... scary? Actually, it was something I heard her say a long time ago. True though, eventually, we all get old enough that you can't plump the raisin back into a grape anymore. If you've set the share, which is a project specific value, and set up the client at all projects 'a' host is attached to, you need to consider that each time you change the home/work/school/default at any project, it will kick in for all projects. Now if you happen to do that change at WCG and hit the update button, obviously, it will pull at the skirts of WCG first. The client ignores the other projects momentarily as their scheduled contact is somewhere in the future. BOINC is keeping count of the amount of work in buffer, the crunching share and the amount of crunching it has done for each, plus the project switch frequency, which also takes account, from 5.8, with checkpoints. All in all whatever the mix is and however big Dolly has grown, after 14 days BOINC will have figured out who should get what and how much. Eventually if you happen to have a short deadline project and it has accumulated enough debt, BOINC will stop pulling work and deplete the time credit. Message is, just let it do its thing, really, and having CPDN and Tanpaku and a number with deadlines in between on a few machines, it does manage. If you attach, remove, change on a daily basis, BOINC will never figure out what you want! I think Didactylos has written it many times: Don't micro-manage. Yes, KISS works well for crunching too. With more than one machine (and especially with the C2Ds) and a decent work queue, you'd have to give up on the rest of life to micro-manage. Best to treat BOINC like a politician and give it time to learn. Precognition is not one of its abilities. ![]() |
||
|
|
|