Can I change the GPU priority to make SETI the highest priority?

Message boards : Number crunching : Can I change the GPU priority to make SETI the highest priority?
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 154
Credit: 2,045,790
RAC: 20,830
United States
Message 2014537 - Posted: 7 Oct 2019, 15:04:23 UTC

I just never noticed this before, but when I've received (or not received) new tasks in my event log it says CPU: job cache full; NVIDIA GPU: not highest priority project. Can I change the GPU priority to make SETI the highest priority?

10/7/2019 8:47:41 AM | Einstein@Home | Not requesting tasks: don't need (CPU: job cache full; NVIDIA GPU: not highest priority project)
10/7/2019 8:47:43 AM | Einstein@Home | Scheduler request completed
10/7/2019 8:49:14 AM | Einstein@Home | Computation for task LATeah1062L06_340.0_0_0.0_4221028_1 finished
10/7/2019 8:49:14 AM | SETI@home | Starting task blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.239_0
10/7/2019 8:49:17 AM | Einstein@Home | Started upload of LATeah1062L06_340.0_0_0.0_4221028_1_0
10/7/2019 8:49:17 AM | Einstein@Home | Started upload of LATeah1062L06_340.0_0_0.0_4221028_1_1
10/7/2019 8:49:18 AM | Einstein@Home | Finished upload of LATeah1062L06_340.0_0_0.0_4221028_1_0
10/7/2019 8:49:18 AM | Einstein@Home | Finished upload of LATeah1062L06_340.0_0_0.0_4221028_1_1
10/7/2019 8:49:18 AM | Milkyway@Home | Sending scheduler request: To report completed tasks.
10/7/2019 8:49:18 AM | Milkyway@Home | Reporting 5 completed tasks
10/7/2019 8:49:18 AM | Milkyway@Home | Not requesting tasks: don't need (CPU: job cache full; NVIDIA GPU: not highest priority project)
10/7/2019 8:49:19 AM | Milkyway@Home | Scheduler request completed
10/7/2019 8:59:38 AM | SETI@home | Computation for task blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.239_0 finished
10/7/2019 8:59:38 AM | Milkyway@Home | Starting task de_modfit_80_bundle4_4s_south4s_bgset_2_1570151403_523665_1
10/7/2019 8:59:40 AM | SETI@home | Started upload of blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.239_0_r1333211256_0
10/7/2019 8:59:42 AM | SETI@home | Finished upload of blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.239_0_r1333211256_0
10/7/2019 9:02:50 AM | Milkyway@Home | Computation for task de_modfit_80_bundle4_4s_south4s_bgset_2_1570151403_523665_1 finished
10/7/2019 9:02:50 AM | Milkyway@Home | Starting task de_modfit_14_bundle4_testing_3s4f_2_1570151403_534178_0
10/7/2019 9:05:20 AM | Milkyway@Home | Computation for task de_modfit_14_bundle4_testing_3s4f_2_1570151403_534178_0 finished
10/7/2019 9:05:20 AM | SETI@home | Starting task blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.181_1
10/7/2019 9:11:21 AM | Einstein@Home | Starting task LATeah0060F_1064.0_976519_0.0_1
10/7/2019 9:17:02 AM | SETI@home | Computation for task blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.181_1 finished
10/7/2019 9:17:02 AM | Milkyway@Home | Starting task de_modfit_80_bundle4_4s_south4s_bgset_2_1570151403_489263_1
10/7/2019 9:17:04 AM | SETI@home | Started upload of blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.181_1_r9258759_0
10/7/2019 9:17:07 AM | SETI@home | Finished upload of blc11_2bit_guppi_58692_08947_HIP84211_0140.1113.0.22.45.181_1_r9258759_0
10/7/2019 9:20:16 AM | Milkyway@Home | Computation for task de_modfit_80_bundle4_4s_south4s_bgset_2_1570151403_489263_1 finished
10/7/2019 9:20:17 AM | Milkyway@Home | Starting task de_modfit_14_bundle4_testing_3s4f_2_1570151403_534165_0
10/7/2019 9:22:49 AM | Milkyway@Home | Computation for task de_modfit_14_bundle4_testing_3s4f_2_1570151403_534165_0 finished
10/7/2019 9:22:49 AM | SETI@home | Starting task blc11_2bit_guppi_58692_08632_HIP84123_0139.1076.818.22.45.199.vlar_0
10/7/2019 9:26:54 AM | SETI@home | Sending scheduler request: To report completed tasks.
10/7/2019 9:26:54 AM | SETI@home | Reporting 3 completed tasks
10/7/2019 9:26:54 AM | SETI@home | Not requesting tasks: don't need (CPU: job cache full; NVIDIA GPU: not highest priority project)
10/7/2019 9:26:55 AM | SETI@home | Scheduler request completed
George

ID: 2014537 · Report as offensive     Reply Quote
Profile Keith Myers Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 10255
Credit: 1,001,745,017
RAC: 1,421,707
United States
Message 2014544 - Posted: 7 Oct 2019, 16:23:43 UTC

"Not the highest priority project" just means that some other work is needed to be done first as in the FIFO order or looming deadline. Or that another project not shown in your log is preempting the use of the gpu. So resource shares matter as far as which project gets worked on the most and likely first, then FIFO order of work received and finally EDF deadlines which trump everything. It all is supposed to balance out in the end over a matter of months.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2014544 · Report as offensive     Reply Quote
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 154
Credit: 2,045,790
RAC: 20,830
United States
Message 2014548 - Posted: 7 Oct 2019, 17:18:25 UTC - in response to Message 2014544.  

"Not the highest priority project" just means that some other work is needed to be done first as in the FIFO order or looming deadline. Or that another project not shown in your log is preempting the use of the gpu. So resource shares matter as far as which project gets worked on the most and likely first, then FIFO order of work received and finally EDF deadlines which trump everything. It all is supposed to balance out in the end over a matter of months.

Once again Keith, I thank you! I was just hoping for a different way of getting SETI more credits. I already have SETI with a Resource Share of 100, and Milkyway & Einstein with 50 each. Maybe I'll set them at 25 each and SETI at 150 and see what happens.
George

ID: 2014548 · Report as offensive     Reply Quote
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 17933
Credit: 408,736,117
RAC: 32,564
United Kingdom
Message 2014551 - Posted: 7 Oct 2019, 17:45:25 UTC

Remember - the resource share for each project is 0 to 1000.
So you could set every project to 1000, and all would get an equal bite at the cherry. In your case I would set SETI to 1000, then scale all the other projects to say 250 or lower each.
Another thing to know is that the overall priority is not a time slice but some strange "deficit share", thus if one project is seen to be lagging behind in its share it will get a temporary bigger than expected share until it has caught up, and the time base for doing so is days not minutes.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 2014551 · Report as offensive     Reply Quote
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 154
Credit: 2,045,790
RAC: 20,830
United States
Message 2014557 - Posted: 7 Oct 2019, 18:43:02 UTC - in response to Message 2014551.  

Thank you Rob. That info is good to know. I'll set my SETI to 1000.
George

ID: 2014557 · Report as offensive     Reply Quote
Profile Keith Myers Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 10255
Credit: 1,001,745,017
RAC: 1,421,707
United States
Message 2014559 - Posted: 7 Oct 2019, 19:04:52 UTC - in response to Message 2014548.  

Once again Keith, I thank you! I was just hoping for a different way of getting SETI more credits. I already have SETI with a Resource Share of 100, and Milkyway & Einstein with 50 each. Maybe I'll set them at 25 each and SETI at 150 and see what happens.

MilkyWay doesn't swamp a machine with work because it only allows a fixed number of tasks on the host at any time. Einstein on the other hand is known "bad actor" which will swamp a machine with work even when it can't finish it all before its short 14 day deadline. The only way to get Einstein under control is assign a different venue for the project on the host and set the days of work allowed to a very small number like 0.1 or 0.2 days and no additional days of work. Einstein is also guilty of not using the normal BOINC APR mechanism and instead relies on the old deprecated task duration factor correction which is wildly inaccurate and overestimates the time it takes to complete a task.

The way to get Seti more credits is to allow it to run more often to the exclusion of other projects like Einstein. I would set SETI resource share to something like 1000 and Einstein to 25 and Milkyway to 50.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2014559 · Report as offensive     Reply Quote
Profile Keith Myers Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 10255
Credit: 1,001,745,017
RAC: 1,421,707
United States
Message 2014560 - Posted: 7 Oct 2019, 19:09:55 UTC - in response to Message 2014551.  
Last modified: 7 Oct 2019, 19:11:41 UTC

overall priority is not a time slice but some strange "deficit share"

The mechanism is called Recent Estimated Credit or REC and is used to balance the books for credit among all projects. This takes the place of what was called the old "debt" mechanism. REC is not normalized for credit either and only works well for projects using CreditNew credit algorithm. So projects like Einstein that eschew CreditNew skew the calculations of REC and allows them to run for much longer than they should. The way to overcome this is with very small caches sizes.

[Edit]There is also a setting in cc_config.xml for the REC mechanism that should be changed from the default 10 days to 1.
<rec_half_life_days>1.000000</rec_half_life_days>
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2014560 · Report as offensive     Reply Quote
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 154
Credit: 2,045,790
RAC: 20,830
United States
Message 2014561 - Posted: 7 Oct 2019, 19:13:03 UTC - in response to Message 2014559.  

MilkyWay doesn't swamp a machine with work because it only allows a fixed number of tasks on the host at any time. Einstein on the other hand is known "bad actor" which will swamp a machine with work even when it can't finish it all before its short 14 day deadline. The only way to get Einstein under control is assign a different venue for the project on the host and set the days of work allowed to a very small number like 0.1 or 0.2 days and no additional days of work. Einstein is also guilty of not using the normal BOINC APR mechanism and instead relies on the old deprecated task duration factor correction which is wildly inaccurate and overestimates the time it takes to complete a task.

If I were to set SETI to "home", Milkyway to "work", and Einstein to "school", would that work to isolate each task where I can separate the amount of work to different amounts for each tasks?
George

ID: 2014561 · Report as offensive     Reply Quote
rob smith Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer moderator
Volunteer tester

Send message
Joined: 7 Mar 03
Posts: 17933
Credit: 408,736,117
RAC: 32,564
United Kingdom
Message 2014563 - Posted: 7 Oct 2019, 19:18:43 UTC

Only if you had different computers in each (virtual) location.
Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?
ID: 2014563 · Report as offensive     Reply Quote
Profile Keith Myers Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 10255
Credit: 1,001,745,017
RAC: 1,421,707
United States
Message 2014565 - Posted: 7 Oct 2019, 20:21:10 UTC - in response to Message 2014563.  

Only if you had different computers in each (virtual) location.

Yes cache size is global unfortunately. So you would have to assign a separate venue to a individual host. But you can then apply those venue settings across all projects. I use a Home venue with a 0.5 day cache for my SBC computers like the Pis and Nanos and run both Einstein and Seti on them. For my hosts that run multiproject concurrently I have to tackle the problem with resource share and a small cache size like 0.5 or 0.1 day. I have the luxury of having a large Seti cache from using a spoofed client so the low daily cache size does not affect me.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2014565 · Report as offensive     Reply Quote
Profile George Project Donor
Volunteer tester
Avatar

Send message
Joined: 23 Oct 17
Posts: 154
Credit: 2,045,790
RAC: 20,830
United States
Message 2014567 - Posted: 7 Oct 2019, 20:45:36 UTC - in response to Message 2014565.  

Just to be clear (and pardon me if my memory is fading), the "venue" is the home/work/school settings? I realize the global settings when set at the online website will set my preferences to one-for-all, but will the "venue" settings on my computer (just one for now) allow me to adjust different settings for each project? If not, will having another computer at the same location (my home) allow me to have different settings for at least my SETI project, and the Milkyway & Einstein projects will have the same settings but different from SETI on my existing computer?

And I'm still waiting, though impatiently, for the release of the 3950X CPU from AMD. That's all I need......
George

ID: 2014567 · Report as offensive     Reply Quote
Profile Keith Myers Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 10255
Credit: 1,001,745,017
RAC: 1,421,707
United States
Message 2014573 - Posted: 7 Oct 2019, 22:14:40 UTC

Yes, "venue" is just another name for location. There are four possible locations. Default/Home/Work/School. If not defined, all hosts get the Default location. You set the location on the Host Details page accessible from the Your Computers list.

On the Seti@home Preferences or Computing Preferences pages, at the top of the page there is a switch labelled "Switch View" which is handy to compare different locations settings on the same page. You can change a location to have different resource share or turn off use of cpu and the daily cache size for example.

Yes you can have completely different locations for different projects. The most common use is to turn off cpu or gpu use for a specific project but you can set different cache sizes too. One thing to look out for is a quirk in BOINC wherever the last Preference change is made on any project is propagated and applied globally across all hosts. So if your primary project is Seti and you want to maintain its preferences, make sure the last preference change you do is at Seti. Just go to any Seti preference page and change one parameter, Save it and then change it back to what is was before. That way the Seti Default preferences is the one in play. You can check which locality preference is in play because it is printed in the startup of the Event Log at every BOINC startup. This has caught out quite a few people, myself included when I was wondering why I wasn't getting any cpu work for example and finally realized I was reading from the Einstein preferences which has no cpu selected.
Seti@Home classic workunits:20,676 CPU time:74,226 hours
ID: 2014573 · Report as offensive     Reply Quote

Message boards : Number crunching : Can I change the GPU priority to make SETI the highest priority?


 
©2019 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.