Message boards :
Number crunching :
Reasource Shareing Between Multiple Projects
Message board moderation
Author | Message |
---|---|
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
Does anyone have any links to information about how to set resource at project@preferences and what variables in BOINC that controls it and how they work? I now have set up a small study of the APR and how it is affected when running multiple projects with different resource set. Since adding Collatz as project number 6 made SETI totally stop from running. I know want to figure out how to set the variables as discribed in project@preferences. resource 100 on 2 tasks gives 50/50 to the projects it is said in the prefs page.. But I don't think it is that simple. It must be more variables to make use of. links to documents that discribe how this work would be appriciated. TRuEQ & TuVaLu |
Wiggo Send message Joined: 24 Jan 00 Posts: 34841 Credit: 261,360,520 RAC: 489 |
Personally I set all my backup projects to 0% resources and leave SETI at 100% as it's my main/preferred project. Also other projects with shorter reporting deadlines will tend to overwhelm SETI with its longer ones so I also have my backup projects set to NNT until I need to call on them and then I select whether I need CPU or GPU from them so I only get enough to keep them happy for a day or so. Cheers. |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
Personally I set all my backup projects to 0% resources and leave SETI at 100% as it's my main/preferred project. Yep, that is also how i've done it. Suspending projects and NNT them... I don't really want to call it a backup project anymore.... Now I want to balance the projects it if possible. To something like: Running 1GPU with 6projects with different credits(I know.....credits). There is 24 hours in 1 day Then I would like to have for example SETI use 12hours and Milkyway 6hours and give the other 4 projects an equal share of what is left of that 24hour period. Atleast I am trying here to figure out a way to come close to something like that. TRuEQ & TuVaLu |
Link Send message Joined: 18 Sep 03 Posts: 834 Credit: 1,807,369 RAC: 0 |
Then I would like to have for example SETI use 12hours and Milkyway 6hours and give the other 4 projects an equal share of what is left of that 24hour period. That would be: SETI: 72 MW: 36 and 4x: 9 |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Personally I set all my backup projects to 0% resources and leave SETI at 100% as it's my main/preferred project. Unfortunately BOINC doesn't function that way. It operates a debt system. So when one project is running it build up a debt for the other(s). If you had 4 projects with equal share. Then BOINC is not going run each project 6 hours a day. It may instead run 1 or 2 at a time and then switch to one, or more, of the others. BOINC user Manual SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
Thank you Hal9000 Unfortunately BOINC doesn't function that way. It operates a debt system. So when one project is running it build up a debt for the other(s). If you had 4 projects with equal share. Then BOINC is not going run each project 6 hours a day. It may instead run 1 or 2 at a time and then switch to one, or more, of the others. That's what I see. 1 or 2 projects switching and then again 1 or 2 projects(different) and then back to the first 2 again. Then I can't really do what I tried to do. Thank you for verifying what I've seen. TRuEQ & TuVaLu |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Thank you Hal9000 It is a function of design. Which on a faster machine seems like it should be able to work in the mode you are wanting. Since the make BOINC to work with as many computers and projects as possible the current mode is better for that scenario. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
Thank you Hal9000 Where is the "debt" stored for the projects? Is there anyway I can see the number and how it is changed for the differnt projects I am running? TRuEQ & TuVaLu |
HAL9000 Send message Joined: 11 Sep 99 Posts: 6534 Credit: 196,805,888 RAC: 57 |
Thank you Hal9000 Like most other information the vales are stored in your client_state.xml. IIRC to see the values in the GUI you select a project, from the projects tab, and select Properties. Then under Scheduling you will see CPU scheduling priority: 0.0 [This is the CPU STD(Short Term Debt)] CPU work fetch priority: 0.0 [This is the CPU LTD(Long Term Debt)] GPU scheduling priority: 0.0 [This is the GPU STD(Short Term Debt)] GPU work fetch priority: 0.0 [This is the GPU LTD(Long Term Debt)] You can compare them to the values in your client_state.xml for verification. If the numbers are 0 then you have no debt for that project/device. Otherwise the numbers will be negative or positive. Your debt, work queue size, and connection setting all play a role in how BOINC decides to download and process tasks. SETI@home classic workunits: 93,865 CPU time: 863,447 hours Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[ |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
IIRC to see the values in the GUI you select a project, from the projects tab, and select Properties. Then under Scheduling you will see CPU scheduling priority: 0.0 [This is the CPU STD(Short Term Debt)] CPU work fetch priority: 0.0 [This is the CPU LTD(Long Term Debt)] GPU scheduling priority: 0.0 [This is the GPU STD(Short Term Debt)] GPU work fetch priority: 0.0 [This is the GPU LTD(Long Term Debt)] There the numbers where. Again thank you. It looks a bit different in 7.0.11 I have under "scheduling" something called scheduling priority that is a different value in each project. It looks to me that it is the project with the lowest number -0.57 that is runing. Then it gets more clear to me which project that will be about next to run. It is SETI that has been waiting to run for about 2 days now. Thank you very much. TRuEQ & TuVaLu |
Claggy Send message Joined: 5 Jul 99 Posts: 4654 Credit: 47,537,079 RAC: 4 |
It looks a bit different in 7.0.11 That's because Boinc 7.0.x uses a completely new method of calculating debt, and has a brand new scheduler too, there's not even a wiki entry on how it works yet, Claggy |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
It looks a bit different in 7.0.11 Ok. No SETI task yet ran. 3days of runtime. It feel like there is a "wanted to be run" factor from 1 or 2 projects when I run 6 at a time on the scheduler... They have a deadline March.... Maybe that's why. If that is part of the calculation then it doesn't seem to matter what resource share I set. :( TRuEQ & TuVaLu |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
I think it would be a neat feature where the BM user could use BM-managing and a sort of relative "Time-sharing(24h-interval)" option in BM to schedule task run times. But that would probebly lead to some of the tasks getting aborted due to deadline. Since some projects sends 100 tasks at a time and other between 1-60 tasks. But I don't think that would be a problem for the user since BM will do it's work based on the "time-sharing(24h-interval)". And it will let the user control how BM would run the scheduling. I dunno if it would be possible.....As always from me, "it's just a thought". TRuEQ & TuVaLu |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
And it looks like I got fooled.... In BM 7.0.11 it seems to be the highest scheduling number that will run next. I think I said the opposite before..... I have 4 projects tha is building "debt" higher then SETI for 3 days now.... They are at aprox -0.4 each and seti is at -0.56 and albert that hasn't been running for 3 days is at -2.22 I wonder how long it will take for the counters to reach albert here... At current rate I would guess 2-4 weeks. TRuEQ & TuVaLu |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
Project preferences in BM 7.x.x observations here. Still 4 projects with aprox -0.4 Albert is -1.87 an improvement And SETI is -0.46 also an improvement. No tasks ran yet. Maybe the projects catch up with SETI tomorrow sometime. TRuEQ & TuVaLu |
Wembley Send message Joined: 16 Sep 09 Posts: 429 Credit: 1,844,293 RAC: 0 |
The more projects you run, the smaller your cache should be. It allows BOINC to react better/faster to debt fluctuations. |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
The more projects you run, the smaller your cache should be. It allows BOINC to react better/faster to debt fluctuations. I have a cache of 2days. TRuEQ & TuVaLu |
LadyL Send message Joined: 14 Sep 11 Posts: 1679 Credit: 5,230,097 RAC: 0 |
I my experience, Boinc 7 is much better at honouring resource share than boinc 6 was. easiest way for a quick overview of debt is to run one cycle of work_fetch_debug logging. Workfetch will be initiated when the calculated remaining work drops below the 'connect every' setting and will then ask for as much work as specified in the 'aditional work' setting. It will ask from the project for which the debt entry is the smallest (calculated as work done recently compared to resource share). If no work is available from that project it will ask from the next larger and so on until it gets work. With SETI this has the usual problem of it being luck whether you get work on a request, so as with boinc 6 you end up getting more work from the other projects. SETI eventually rises in priority until you get lucky. With BOINC 7 cache needs to be set quite differently from the approach used with previous versions to get the same effect. Because it will wait until the 'connect every' minimum level is reached before it asks for work (instead of the frequent top ups with previous) if you want to be sure to have work for 3 days available at all times on the machine you need to set 'connect every' to 3. If you want it to ask often for work (to get that elusive SETI task) you should set 'additional' to a small value. |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
I my experience, Boinc 7 is much better at honouring resource share than boinc 6 was. my scheduling priority number for projects that runs seems to be between -0.4 -- -0.5 SETI went below -0.46 today and did 6 tasks and now it is running from time to time. Now all I wait for is Albert that has -2.1 to start to run with the other projects that runs when they are higher then -0.46(aprox). I am happy to see my SETI caught up in the debt calculation. Is there any fast way to reset the number -2.1 to get my albert to join the other projects instead of the long time waiting? And yes I think this way feels ok for running alot of projects. A bit slow with my SETI catching up and now Albert too. Collatz, Milkyway, Primegrid that I run rotates nicely and i think SETI will now as well. TRuEQ & TuVaLu |
TRuEQ & TuVaLu Send message Joined: 4 Oct 99 Posts: 505 Credit: 69,523,653 RAC: 10 |
I have it set to 0 (always connected).
How do I run 1 cycle of work_fetch_debug logging? TRuEQ & TuVaLu |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.