I have what I think is a bug. I have a climate prediction task that estimates it will complete 2 hours before he deadline. I have boinc set so that no new tasks will be downloaded from any of my connected projects.

There are many SZTAKI tasks to be done (about 15) and 2 climate prediction tasks.

the laptop has a quad core. I have to suspend all the SZTAKI tasks except two in order to let the climate prediction tasks run. 1 climate prediction task has 12 days left and about 4 weeks until dead line. This one is running at high priority. The 2nd climate prediction task will finish 2 hours before the dead line at the current pace and it is not running at high priority. This 2nd climate prediction task does report a dead line warning.

If I have more than 3 SZTAKI tasks running, the 2nd cllimate prediction task goes to "waiting to run" state.
I expect the 2nd climate prediction should be running at high priority.
Why isn't it?

Here is the beginning of my output. I just rebooted so it is small. ugh and poorly formatted. This was copied from BoincView.


Host Project Date Message
bigLaptop --- 7/19/2013 8:55:21 PM No config file found - using defaults
bigLaptop --- 7/19/2013 8:55:21 PM Starting BOINC client version 7.0.64 for windows_x86_64
bigLaptop --- 7/19/2013 8:55:21 PM log flags: file_xfer, sched_ops, task
bigLaptop --- 7/19/2013 8:55:21 PM Libraries: libcurl/7.25.0 OpenSSL/1.0.1 zlib/1.2.6
bigLaptop --- 7/19/2013 8:55:21 PM Running as a daemon
bigLaptop --- 7/19/2013 8:55:21 PM Data directory: C:\ProgramData\BOINC
bigLaptop --- 7/19/2013 8:55:21 PM Running under account boinc_master
bigLaptop --- 7/19/2013 8:55:21 PM Processor: 4 GenuineIntel Intel(R) Core(TM) i3 CPU M 330 @ 2.13GHz [Family 6 Model 37 Stepping 2]
bigLaptop --- 7/19/2013 8:55:21 PM 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 cx16 sse4_1 sse4_2 popcnt syscall lm vmx tm2 pbe
bigLaptop --- 7/19/2013 8:55:21 PM OS: Microsoft Windows 7: Home Premium x64 Edition, Service Pack 1, (06.01.7601.00)
bigLaptop --- 7/19/2013 8:55:21 PM Memory: 3.68 GB physical, 7.36 GB virtual
bigLaptop --- 7/19/2013 8:55:21 PM Disk: 453.66 GB total, 401.76 GB free
bigLaptop --- 7/19/2013 8:55:21 PM Local time is UTC -7 hours
bigLaptop --- 7/19/2013 8:55:21 PM No usable GPUs found
bigLaptop --- 7/19/2013 8:55:21 PM Config: GUI RPCs allowed from:
bigLaptop --- 7/19/2013 8:55:21 PM princess
bigLaptop rosetta@home 7/19/2013 8:55:21 PM URL http://boinc.bakerlab.org/rosetta/; Computer ID 1489611; resource share 5
bigLaptop boincsimap 7/19/2013 8:55:21 PM URL http://boincsimap.org/boincsimap/; Computer ID 210517; resource share 5
bigLaptop climateprediction.net 7/19/2013 8:55:21 PM URL http://climateprediction.net/; Computer ID 1130955; resource share 75
bigLaptop Einstein@Home 7/19/2013 8:55:21 PM URL http://einstein.phys.uwm.edu/; Computer ID 3892017; resource share 75
bigLaptop LHC@home 1.0 7/19/2013 8:55:21 PM URL http://lhcathomeclassic.cern.ch/sixtrack/; Computer ID 9938902; resource share 5
bigLaptop Milkyway@Home 7/19/2013 8:55:21 PM URL http://milkyway.cs.rpi.edu/milkyway/; Computer ID 430394; resource share 5
bigLaptop SETI@home 7/19/2013 8:55:21 PM URL http://setiathome.berkeley.edu/; Computer ID 5804106; resource share 75
bigLaptop SZTAKI Desktop Grid 7/19/2013 8:55:21 PM URL http://szdg.lpds.sztaki.hu/szdg/; Computer ID 337479; resource share 750
bigLaptop SETI@home 7/19/2013 8:55:21 PM General prefs: from SETI@home (last modified 07-Jul-2013 20:30:15)
bigLaptop SETI@home 7/19/2013 8:55:21 PM Computer location: work
bigLaptop SETI@home 7/19/2013 8:55:21 PM General prefs: no separate prefs for work; using your defaults
bigLaptop --- 7/19/2013 8:55:21 PM Preferences:
bigLaptop --- 7/19/2013 8:55:21 PM max memory usage when active: 1883.39MB
bigLaptop --- 7/19/2013 8:55:21 PM max memory usage when idle: 3390.10MB
bigLaptop --- 7/19/2013 8:55:21 PM max disk usage: 100.00GB
bigLaptop --- 7/19/2013 8:55:21 PM don't compute while active
bigLaptop --- 7/19/2013 8:55:21 PM don't use GPU while active
bigLaptop --- 7/19/2013 8:55:21 PM suspend work if non-BOINC CPU load exceeds 26 %
bigLaptop --- 7/19/2013 8:55:21 PM (to change preferences, visit a project web site or select Preferences in the Manager)
bigLaptop --- 7/19/2013 8:55:21 PM Not using a proxy
bigLaptop --- 7/19/2013 8:55:22 PM Running CPU benchmarks
bigLaptop --- 7/19/2013 8:55:22 PM Suspending computation - CPU benchmarks in progress
bigLaptop --- 7/19/2013 8:55:54 PM Benchmark results:
bigLaptop --- 7/19/2013 8:55:54 PM Number of CPUs: 4
bigLaptop --- 7/19/2013 8:55:54 PM 1924 floating point MIPS (Whetstone) per CPU
bigLaptop --- 7/19/2013 8:55:54 PM 6543 integer MIPS (Dhrystone) per CPU
bigLaptop climateprediction.net 7/19/2013 8:55:55 PM Restarting task hadcm3n_o3je_1940_40_008382822_0 using hadcm3n version 607 in slot 4
bigLaptop climateprediction.net 7/19/2013 8:55:55 PM Restarting task hadcm3n_o3nx_1980_40_008384805_0 using hadcm3n version 607 in slot 1
bigLaptop SZTAKI Desktop Grid 7/19/2013 8:55:55 PM Restarting task aaaa0b60-d41c-41ea-8e5f-d1777c9dfd53_64a779d0-07b0-4c17-b6bb-5a6c08258378_359454_1 using search version 206 in slot 0
bigLaptop SZTAKI Desktop Grid 7/19/2013 8:55:55 PM Restarting task aaaa0b60-d41c-41ea-8e5f-d1777c9dfd53_4bc442ff-b5a7-4331-bef4-0fa2911216c3_359477_1 using search version 206 in slot 2

I have no answer, but have one CPDN running along with two CEP2 jobs on three cores of an i5-3550. I had thought that the jobs ran in the order in which they were downloaded (except in case a deadline has to be met), but in fact if I don't limit the CEP2 jobs to only two running at a time with an app_config file, then three of them will run and the CPDN will sit idle, even though it is by far the oldest, having been downloaded 9 days ago.

I have never really tried figuring out BOINC's scheduling (now 7.2.5 x64 under Win7 64-bit for me, but probably 7.0.64 when the CPDN was downloaded), but good luck in finding an answer, if there is one.

To see if it can run all work by deadline, make a cc_config.xml file in Notepad, save it into your BOINC Data directory (C:\ProgramData\BOINC) and add into it:

<cc_config>
   <log_flags>
       <cpu_sched_debug>1</cpu_sched_debug>
       <rr_simulation>1</rr_simulation>
       <sched_op_debug>1</sched_op_debug>
    </log_flags>
  <options>
  </options>
</cc_config> 

When saving it, make sure to make it an 'all files' option, ANSI format. After saving, double check that the file extension is only .xml, not also .txt. If it is also .txt, rename it so the extension is only .xml

Next open BOINC Manager, Advanced, Read config file(s).
The above debug flags will now add a lot more information into your Event Log.
Run for two or three iterations, then suspend BOINC, and change the lines in cc_config.xml to:

<cc_config>
   <log_flags>
       <cpu_sched_debug>0</cpu_sched_debug>
       <rr_simulation>0</rr_simulation>
       <sched_op_debug>0</sched_op_debug>
    </log_flags>
  <options>
  </options>
</cc_config> 

Save changes to the file.
Next open BOINC Manager, Advanced, Read config file(s).
You can now resume BOINC.

Then open the Event Log (CTRL+SHIFT+E), copy the two or three iterations of the debug flags from the log and post that here.

---

I have no answer, but have one CPDN running along with two CEP2 jobs on three cores of an i5-3550. I had thought that the jobs ran in the order in which they were downloaded (except in case a deadline has to be met), but in fact if I don't limit the CEP2 jobs to only two running at a time with an app_config file, then three of them will run and the CPDN will sit idle, even though it is by far the oldest, having been downloaded 9 days ago.

I have never really tried figuring out BOINC's scheduling (now 7.2.5 x64 under Win7 64-bit for me, but probably 7.0.64 when the CPDN was downloaded), but good luck in finding an answer, if there is one.

You forget about timeshare which is tracked. In principle tasks run FIFO -per project- and EDF if any task is under deadline threat. Since the CPDN is not under deadline threat, I'm sure, normal alternation between projects applies, which is default I think 60 minutes [the switch application you can set to influence how long a task runs once started... not quite but it helps to complete tasks in least interruption]. Of course with CEP2, you have LAIM set, else if EDF kicks in, interrupted tasks get unloaded before the checkpoint is written, which is the normal switch application moment when time's up. With the large checkpoint intervals, particular at #2 and #11 (3rd and 12th really), you could be loosing hours progress. Of course that is then predicated on having enough memory, but normally if tasks are held in memory, the OS will eventually move them whole or part into virtual memory, if that's active and allowed to be used by BOINC [recommended].

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

You forget about timeshare which is tracked. In principle tasks run FIFO -per project- and EDF if any task is under deadline threat. Since the CPDN is not under deadline threat, I'm sure, normal alternation between projects applies, which is default I think 60 minutes [the switch application you can set to influence how long a task runs once started... not quite but it helps to complete tasks in least interruption]. Of course with CEP2, you have LAIM set, else if EDF kicks in, interrupted tasks get unloaded before the checkpoint is written, which is the normal switch application moment when time's up. With the large checkpoint intervals, particular at #2 and #11 (3rd and 12th really), you could be loosing hours progress. Of course that is then predicated on having enough memory, but normally if tasks are held in memory, the OS will eventually move them whole or part into virtual memory, if that's active and allowed to be used by BOINC [recommended].

All good points, except that it seems to be that CEP2 always takes precedence, and I don't ever recall seeing the 60-minute switch. Yes, I always have LAIM set, but even more I run the BOINC folder out of Ramdisk (12 GB, with usually half of it free), so whatever happens should be going fast.

But fate has intervened; CPDN is out of work, and so I am going to CEP2 exclusively on that machine; problem solved.

In reading through this explanation by Claggy, it appears that the behavior I observe is probably due to the Recent Estimated Credit (REC):
http://boinc.berkeley.edu/dev/forum_thread.php?id=8513&postid=49939#49939

That is, since I had been running CPDN for several days on all three cores without any CEP2 jobs at all, it had built up a large REC for CPDN. Then, when I added the CEP2 work, it preferred that to the CPDN work. Eventually, it would have equalized out at the resource share I had chosen (equal for both projects in my case).

That is all very interesting, but beyond human comprehension in most cases. I think it would be better if BOINC just asked which projects you want to run on which cores (with options for secondary projects if the primary ones are out of work), and ditch the whole REC scheme, and whatever else they have. Fortunately, they did not ask my advice, since it may overlook something important. Or it may just be that they are trying to over-engineer the whole thing.

Or you could stop watching the BOINC [which it hates] and let it run ;>)

---

Coelum Non Animum Mutant, Qui Trans Mare Currunt

Or you could stop watching the BOINC [which it hates] and let it run ;>)

That has occurred to me. But then, what is the fun in it?

Perhaps more to the point, you don't know that your selections are working properly, since you get no feedback until days, if not weeks later. And then you may discover that something was misconfigured all along. I think it is just another attempt to "make things easy for you" that in fact introduces unnecessary complications. If this project isn't for techies, then who is it for? The average user won't bother with so many projects anyway (or at least shouldn't).