Resource management

Message boards : Questions and problems : Resource management
Message board moderation

To post messages, you must log in.

AuthorMessage
Pete

Send message
Joined: 15 Feb 15
Posts: 1
United Kingdom
Message 60321 - Posted: 15 Feb 2015, 11:28:55 UTC

Hi, I am having a problem with manageing the resources that Boinc assigns to the 2 projects I run. Whatever I do Boinc seems to allocate about equal or more running time to seti. I want to allocate about 70-30 to einstein. I have a gpu which is only allowed to run Einstein, this happens successfully.The E@h project only downloads a short queue but there is always at least one work unit available. The S@h project seems to want a much longer queue. The subjective view I have is that whenever I glance at the task queues is that S@h is happily running most tasks,occasionally E@h is running some. I have tweaked the resource allocation such that currently E@h is 200 and S@h is 40. I even had S@h down to 10 at one point and it seemed to make a small difference but not what you would expect with a 200-10 distribution. I am currently trying to force the allocation I want by manually suspending tasks but this is not the way to go obviously.
The initial part of the log is as below.
15/02/2015 09:50:52 | | cc_config.xml not found - using defaults
15/02/2015 09:50:52 | | Starting BOINC client version 7.4.36 for windows_x86_64
15/02/2015 09:50:52 | | log flags: file_xfer, sched_ops, task
15/02/2015 09:50:52 | | Libraries: libcurl/7.39.0 OpenSSL/1.0.1j zlib/1.2.8
15/02/2015 09:50:52 | | Data directory: C:\ProgramData\BOINC
15/02/2015 09:50:52 | | Running under account peter
15/02/2015 09:50:52 | | CUDA: NVIDIA GPU 0: GeForce GTX 650 (driver version 347.25, CUDA version 7.0, compute capability 3.0, 2048MB, 1944MB available, 813 GFLOPS peak)
15/02/2015 09:50:52 | | OpenCL: NVIDIA GPU 0: GeForce GTX 650 (driver version 347.25, device version OpenCL 1.1 CUDA, 2048MB, 1944MB available, 813 GFLOPS peak)
15/02/2015 09:50:52 | | OpenCL CPU: Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz (OpenCL driver vendor: Intel(R) Corporation, driver version 4.2.0.99, device version OpenCL 1.2 (Build 99))
15/02/2015 09:50:52 | | app version refers to missing GPU type intel_gpu
15/02/2015 09:50:52 | SETI@home | Application uses missing intel_gpu GPU
15/02/2015 09:50:52 | | app version refers to missing GPU type intel_gpu
15/02/2015 09:50:52 | SETI@home | Application uses missing intel_gpu GPU
15/02/2015 09:50:52 | | app version refers to missing GPU type intel_gpu
15/02/2015 09:50:52 | Einstein@Home | Application uses missing intel_gpu GPU
15/02/2015 09:50:52 | | Host name: Peter
15/02/2015 09:50:52 | | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-4770K CPU @ 3.50GHz [Family 6 Model 60 Stepping 3]
15/02/2015 09:50:52 | | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 fma cx16 sse4_1 sse4_2 movebe popcnt aes f16c rdrandsyscall nx lm avx avx2 vmx tm2 pbe fsgsbase bmi1 smep bmi2
15/02/2015 09:50:52 | | OS: Microsoft Windows 8.1: Professional x64 Edition, (06.03.9600.00)
15/02/2015 09:50:52 | | Memory: 7.94 GB physical, 9.19 GB virtual
15/02/2015 09:50:52 | | Disk: 223.05 GB total, 193.05 GB free
15/02/2015 09:50:52 | | Local time is UTC +0 hours
15/02/2015 09:50:52 | Einstein@Home | URL http://einstein.phys.uwm.edu/; Computer ID 11715962; resource share 200
15/02/2015 09:50:52 | SETI@home | URL http://setiathome.berkeley.edu/; Computer ID 7472948; resource share 40
15/02/2015 09:50:52 | SETI@home | General prefs: from SETI@home (last modified 06-Feb-2015 12:13:52)
15/02/2015 09:50:52 | SETI@home | Computer location: home
15/02/2015 09:50:52 | SETI@home | General prefs: no separate prefs for home; using your defaults
15/02/2015 09:50:52 | | Reading preferences override file
15/02/2015 09:50:52 | | Preferences:
15/02/2015 09:50:52 | | max memory usage when active: 5691.32MB
15/02/2015 09:50:52 | | max memory usage when idle: 7317.41MB
15/02/2015 09:50:52 | | max disk usage: 193.49GB
15/02/2015 09:50:52 | | suspend work if non-BOINC CPU load exceeds 50%
15/02/2015 09:50:52 | | (to change preferences, visit a project web site or select Preferences in the Manager)
15/02/2015 09:50:52 | | Not using a proxy
15/02/2015 10:04:30 | SETI@home | Computation for task 14fe13ab.21002.2521.438086664203.12.213_1 finished
15/02/2015 10:04:33 | SETI@home | Started upload of 14fe13ab.21002.2521.438086664203.12.213_1_0
15/02/2015 10:04:36 | SETI@home | Finished upload of 14fe13ab.21002.2521.438086664203.12.213_1_0
15/02/2015 10:07:44 | SETI@home | update requested by user
15/02/2015 10:07:46 | SETI@home | Sending scheduler request: Requested by user.
15/02/2015 10:07:46 | SETI@home | Reporting 1 completed tasks
15/02/2015 10:07:46 | SETI@home | Not requesting tasks: some task is suspended via Manager
15/02/2015 10:07:47 | SETI@home | Scheduler request completed
15/02/2015 10:08:12 | SETI@home | task 30au12ab.31305.12750.438086664204.12.79_0 resumed by user
15/02/2015 10:33:06 | SETI@home | task 14fe13ab.7776.476.438086664204.12.1_1 resumed by user
15/02/2015 10:49:39 | SETI@home | Computation for task 04my12ad.8234.2686.438086664195.12.190.vlar_1 finished
15/02/2015 10:49:39 | SETI@home | Computation for task 30au12ab.31305.12750.438086664204.12.79_0 finished
15/02/2015 10:49:41 | SETI@home | Started upload of 04my12ad.8234.2686.438086664195.12.190.vlar_1_0
15/02/2015 10:49:41 | SETI@home | Started upload of 30au12ab.31305.12750.438086664204.12.79_0_0
15/02/2015 10:49:45 | SETI@home | Finished upload of 04my12ad.8234.2686.438086664195.12.190.vlar_1_0
15/02/2015 10:49:45 | SETI@home | Finished upload of 30au12ab.31305.12750.438086664204.12.79_0_0
15/02/2015 10:50:42 | SETI@home | task 14fe13ab.7776.476.438086664204.12.43_0 resumed by user
15/02/2015 10:52:24 | Einstein@Home | Computation for task PB0092_02781_24_0 finished
15/02/2015 10:52:24 | Einstein@Home | Starting task p2030.20131012.G176.45-00.18.S.b6s0g0.00000_2176_1
15/02/2015 10:52:26 | Einstein@Home | Started upload of PB0092_02781_24_0_0
15/02/2015 10:52:26 | Einstein@Home | Started upload of PB0092_02781_24_0_1
15/02/2015 10:52:27 | Einstein@Home | Finished upload of PB0092_02781_24_0_0
15/02/2015 10:52:27 | Einstein@Home | Finished upload of PB0092_02781_24_0_1
15/02/2015 11:03:58 | SETI@home | Computation for task 28no12ad.7033.130605.438086664204.12.173_1 finished
15/02/2015 11:04:00 | SETI@home | Started upload of 28no12ad.7033.130605.438086664204.12.173_1_0
15/02/2015 11:04:04 | SETI@home | Finished upload of 28no12ad.7033.130605.438086664204.12.173_1_0
ID: 60321 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 60341 - Posted: 16 Feb 2015, 14:39:45 UTC - in response to Message 60321.  
Last modified: 16 Feb 2015, 14:41:06 UTC

Changing resource share and expecting BOINC to follow it takes time. Not as in 'tomorrow it will do what I tell it to', but serious time, multiple weeks, months even.
By constantly interfering, by changing resource shares wildly or suspending tasks of one project, you're making it more difficult for BOINC to try to do what you want it to do.

You will need to give BOINC time to learn how long various tasks of a project take, this can take a minimum of 11 faultless tasks, but probably more. Then both projects should have constant up-time.

It can happen that when project A is highest priority and asks for work, but its servers are down, that BOINC will then ask project B for work and load up the cache full with work for that project. Even if that project is the lowest in resource share. Next BOINC will first have to run through that cache, to get back to the low-water mark and ask for work again. Is project A still down, it'll fill up on project B again.

It may be that between these touch-ups, project A is up. BOINC will only ask for work when it reaches the minimum water mark, set by "Store at least N days of work". Normally, when both projects are up, the cache of BOINC v7 is composed of "Store at least N days of work + Store an additional N days of work" for 50% of project A, 50% of project B.

So it would be prudent to set a low cache in the days you want to have BOINC work out the resource share settings you tell it to use. Setting 10 + 10 days is just counterproductive. Set low cache, e.g. 0 + 0.1 days and keep an eye on BOINC. Do NOT interfere. That is also counterproductive. Keep an eye on the projects, do they have down time, do they have the work you tell BOINC to request?

And then in a month, see how things go. Probably a lot better than they go now.
BOINC will really do what you want it to do, as long as you can muster up the patience for it to learn.
ID: 60341 · Report as offensive
Cruncher Pete

Send message
Joined: 16 Oct 10
Posts: 27
Australia
Message 60346 - Posted: 17 Feb 2015, 0:22:19 UTC

Pete, I sympathies with you for the same reason I now only run just one project at a time until I achieve my target or get sick of running it. Jord was doing a good job in explaining the reasons why this is so. Having said that, I think it is time the Developers are made aware of this. It is an option that is not an option for it does not do what we expect it to do. It is not good enough to expect that it might take multiply weeks before your selected option takes effect. It seems to me that the option is useless for it is controlled and effected by your cache. For example the lowest cache setting is not suitable for me for I wish to run my project and a backup one should the project develop a problem for at least three days so that I can go away from home on a long weekend without worrying that my machines will waste electricity if the project is down. Clearly the option to have the maximum of 10 days cache would result in an extraordinary time before BOINC sorts itself out and balances the two project to your wishes.

My recommendation is that you either run just one CPU project at a time or write to the developers or Dr David Anderson your concern. The more feedback they receive from users the more likely they will look at the problem, for I too consider it to be that the option is not what one expects should you wish to use it as either its use is not well explained or having to wait for a month before it might take effect is useless..
ID: 60346 · Report as offensive

Message boards : Questions and problems : Resource management

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.