For some reason actual scheduling has changed from the specified 50% SETI and 50% Climate

Message:
"4/6/2006 7:58:00 PM||Using earliest-deadline-first scheduling because computer is overcommitted."


But SETI jobs sum of "to complete" times (few hours)are a trivial fraction of time till "Report Deadline" (4/18/2006). Yet they are continually preempting the Climate job. Apparently because it has a Report Deadline of 2/27/2007 since it still has 3513 hr to complete.

Why?

---

Please post the messages in front of the "earliest deadline first" message as well, as that will tell which project it will go into EDF mode for. And if at that time the Seti unit is still crunching, it'll be changing to EDF mode at the next switch time.

Also post your connect to server time plus how many work units of both the projects you have in queue.

More about EDF in the Wiki.

---

You need to read up of it, as there are numerous parameters that effect this.....

---

BOINC Wiki

messages in front of the "earliest deadline first" message

4/6/2006 7:56:56 PM||Starting BOINC client version 5.2.13 for windows_intelx86
4/6/2006 7:56:56 PM||libcurl/7.14.0 OpenSSL/0.9.8 zlib/1.2.3
4/6/2006 7:56:56 PM||Data directory: C:\\Program Files\\BOINC
4/6/2006 7:56:56 PM||Processor: 2 GenuineIntel Intel(R) Pentium(R) 4 CPU 2.60GHz
4/6/2006 7:56:56 PM||Memory: 510.98 MB physical, 1.22 GB virtual
4/6/2006 7:56:56 PM||Disk: 43.23 GB total, 25.23 GB free
4/6/2006 7:56:56 PM|SETI@home|Computer ID: 783770; location: home; project prefs: default
4/6/2006 7:56:56 PM|climateprediction.net|Computer ID: 305492; location: home; project prefs: default
4/6/2006 7:56:56 PM||General prefs: from climateprediction.net (last modified 2005-12-11 12:15:10)
4/6/2006 7:56:56 PM||General prefs: no separate prefs for home; using your defaults
4/6/2006 7:56:57 PM||Remote control not allowed; using loopback address
4/6/2006 7:56:59 PM||Running CPU benchmarks
4/6/2006 7:57:58 PM||Benchmark results:
4/6/2006 7:57:58 PM|| Number of CPUs: 2
4/6/2006 7:57:58 PM|| 1144 double precision MIPS (Whetstone) per CPU
4/6/2006 7:57:58 PM|| 999 integer MIPS (Dhrystone) per CPU
4/6/2006 7:57:58 PM||Finished CPU benchmarks
4/6/2006 7:58:00 PM|climateprediction.net|Deferring computation for result hadcm3lb_4jzt_05000517_0
4/6/2006 7:58:00 PM|SETI@home|Resuming computation for result 05au01ab.28341.28529.342332.1.91_0 using setiathome version 418
4/6/2006 7:58:00 PM|SETI@home|Resuming computation for result 30my02aa.9853.18002.136088.1.122_1 using setiathome version 418
4/6/2006 7:58:00 PM||Resuming computation and network activity
4/6/2006 7:58:00 PM||request_reschedule_cpus: Resuming activities
4/6/2006 7:58:00 PM||Suspending work fetch because computer is overcommitted.
4/6/2006 7:58:00 PM||Using earliest-deadline-first scheduling because computer is overcommitted.
4/6/2006 8:03:52 PM||request_reschedule_cpus: process exited

The SETI Que has now worked its way down to about 5 hr in 2 jobs.
(both jobs are active, I have dual CPUs) No new SETI jobs have been downloaded in the last day or so.
The inactive Climate job has 498 hr completed and 3514 to go.
And

---

It appears that the CPDN task has deadline problems. This can lead to a problem on multi-CPU machines since you run out of work for one CPU and download a bunch of work that will preempt the CPDN task.

First solution is to reduce your queue size so that only one task from another project will download at a time. This way CPDN will keep one CPU when new work is downloaded.

Second solution is to suspend (or no new work) all projects except CPDN. This will get a model for each CPU and hopefully by the time one finishes the scheduler will have been fixed.

Also since the CPDN project is not terribly concerned about deadlines you can just let it run as is until a fix for the scheduler is released.

---

BOINC WIKI

BOINCing since 2002/12/8

It appears that the CPDN task has deadline problems. This can lead to a problem on multi-CPU machines since you run out of work for one CPU and download a bunch of work that will preempt the CPDN task.

First solution is to reduce your queue size so that only one task from another project will download at a time. This way CPDN will keep one CPU when new work is downloaded.

Second solution is to suspend (or no new work) all projects except CPDN. This will get a model for each CPU and hopefully by the time one finishes the scheduler will have been fixed.

Also since the CPDN project is not terribly concerned about deadlines you can just let it run as is until a fix for the scheduler is released.

---

Thanks for the reply.
The first suggestion appears best since the 2nd will get two Climate and no SETI.
But how do I adjust que size? I haven't found that any where.

---

Thanks for the reply.
The first suggestion appears best since the 2nd will get two Climate and no SETI.
But how do I adjust que size? I haven't found that any where.

It's the connect to network about every ___ setting under your general preferences.

Seti: Your Account, View or Edit General Preferences, edit them with the link and do not forget to save the changes to the website with the button.

Or CPDN:
Your Account, View or Edit General Preferences, same thing as above.

When saved, open Boinc Manager, Projects tab, select the Seti@Home or CPDN project by clicking on it, press Update on the left.

Changes made on one project will propagate to the other projects, so you only need to change preferences on one site.

---

You may also try to set CPDN to no new work but to a higher share (99%). This will cause CPDN to hardly ever get suspended but will also force BOINC to download only Seti-WUs if the cache is empty.

---