Now that proteinfolding is delivering work units again, I find that the scheduling done by the 4.13 boinc client works funny. I have two XEON hyperthreaded CPUs, 4 Gig RAM, and stuff. Now when the machine is not too busy, it does not run all the processes it could. I have lots of work units for setiathome, two work units for climateprediction, and one work unit for proteinfolding. Yet it schedules only three processes that actually run most of the time, and it seems to be making scheduling decisions more often than once an hour. Here is the "top" command output, showing only boinc processes. Notice that other worker processes are in the process table, but they do not run.

It seems that when climateprediction stops a process, it does it in a way that it appears in the "Stopped" count in "top", but the others do not seem to work that way.

Is something wrong, or is this how it is supposed to work?

18:36:32 up 2 days, 7:55, 3 users, load average: 3.31, 3.31, 4.81
146 processes: 141 sleeping, 5 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 8.0% 298.8% 4.8% 0.4% 0.0% 0.0% 86.8%
cpu00 3.2% 56.1% 2.2% 0.5% 0.0% 0.0% 37.9%
cpu01 4.4% 46.8% 2.0% 0.0% 0.0% 0.3% 46.3%
cpu02 0.1% 99.4% 0.3% 0.0% 0.0% 0.0% 0.0%
cpu03 0.1% 96.5% 0.3% 0.1% 0.0% 0.0% 2.7%
Mem: 4100860k av, 3081084k used, 1019776k free, 0k shrd, 114860k buff
967924k actv, 1511584k in_d, 69508k in_c
Swap: 8193076k av, 22724k used, 8170352k free 2447072k cached

PID PPID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM CTIME COMMAND
17745 31134 boinc 39 19 15500 15M 1292 R N 101.2 0.3 126:14 setiathome_4.02_i686-pc-linux-gnu
31139 31136 boinc 39 19 47008 45M 4000 R N 99.7 1.1 1275m hadsm3um_4.04_i686-pc-linux-gnu 23975 31136
31140 31135 boinc 39 19 47016 45M 4000 R N 99.5 1.1 1325m hadsm3um_4.04_i686-pc-linux-gnu 26470 31135
31134 1 boinc 15 0 2320 2320 1248 S 0.0 0.0 1734m /boinc/boinc_4.13_i686-pc-linux-gnu
31135 31134 boinc 34 19 908 908 708 S N 0.0 0.0 0:05 hadsm3_4.04_i686-pc-linux-gnu 2qzd_100149577
31136 31134 boinc 34 19 908 908 708 S N 0.0 0.0 0:04 hadsm3_4.04_i686-pc-linux-gnu 2s01_100150910
14608 31134 boinc 34 19 43372 42M 596 S N 0.0 1.0 145:04 mfoldB120_4.20_i686-pc-linux-gnu
15138 31134 boinc 34 19 15356 12M 1300 S N 0.0 0.3 32:25 setiathome_4.02_i686-pc-linux-gnu
15141 31134 boinc 34 19 15116 14M 1184 S N 0.0 0.3 246:04 setiathome_4.02_i686-pc-linux-gnu
15142 31134 boinc 34 19 15352 1528 1288 S N 0.0 0.0 11:49 setiathome_4.02_i686-pc-linux-gnu

---

I also had some problems with HT until I switched the "keep process in memory" option off.

Regards Hans

P.S:

Unloading/Reloading the processes when switching only costs a negligible amount of time.

---

> I also had some problems with HT until I switched the "keep process in memory"
> option off.

What was the nature of the problems you had?
>
> Regards Hans
>
> P.S:
>
> Unloading/Reloading the processes when switching only costs a negligible
> amount of time.
>
Loading and unloading processes may not take long, but it is my understanding that the trouble is that with climate-prediction, you can lose up to a whole trickle if you just stop a process, and that can cost you a large part of a day in computation.

---