15:05:48 <skalnik1> #startmeeting Power Management for Fedora
15:05:48 <zodbot_> Meeting started Thu Sep 10 15:05:48 2009 UTC.  The chair is skalnik1. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:05:48 <zodbot_> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:06:05 <skalnik1> Hello PM funs
15:06:33 <skalnik1> I'm sorry I'm a bit late and started the meeting on false channel :-(
15:07:31 <skalnik1> Unfortunately, I'm alone from PM meeting on fedora-meeting
15:08:22 <skalnik1> because there is Developer Conference 2009 in Brno and my colleagues are busy with this event.
15:08:41 <skalnik1> Phill Knirsch is on vacation
15:09:56 <skalnik1> Therefore I have no topics to discuss
15:10:52 <skalnik1> I'd like to inform you shortly about just running DevCon in Brno
15:12:21 <skalnik1> Power Management is quite active. We presented two presentations and There is just finishing Watt hunting competition.
15:13:27 <skalnik1> We spoke about our work and results.
15:14:07 <skalnik1> We informed audience about BLTK, tuned and scomes packages
15:14:31 <mjg59> skalnik1: Are there available progress reports on these?
15:15:26 <skalnik1> mjg59: changes in bltk are just discussed with upstream
15:16:22 <skalnik1> tuned package was extended about tuned-adm that switches predefined profiles
15:16:45 <skalnik1> we add new info on our SIG pages
15:17:15 <skalnik1> but I suppose We should focus on improvement these pages
15:17:38 <mjg59> The results of the watt hunting competition would be interesting
15:18:07 <skalnik1> ah, sorry you mentioned watt hunting
15:20:05 <skalnik1> mjg59:  I will ask colleagues for insertion results of the competition on the web sites
15:21:10 <skalnik1> complete info, it could be interesting to have more competitor
15:21:20 <mjg59> That would be great
15:21:29 <skalnik1> s/competitor/competitors/ ;-)
15:22:54 <mgahagan> How are you guys measuring power consumption... external watt meter or just going by ACPI batter life estimates?
15:23:58 <skalnik1> mgahagan: we used three ways to measure power consumption
15:24:32 <skalnik1> two direct - battery info as wrote mgahagan (this uses bltk)
15:24:56 <skalnik1> then external watt meter but this was only for first imagines
15:25:52 <skalnik1> because I discussed socket watt meters accuracy with university people here in brno
15:27:02 <skalnik1> and the answer was these watt meters are not usable
15:27:16 <mgahagan> skalnik1, ahh ok.. was just wondering.... I know accurate watt meters for AC power get very, very expensive.
15:27:59 <skalnik1> mgahagan: yes, the cheapest accurate watt meters starts on aprox. 1000 $
15:28:44 <skalnik1> but common price can be 30k - 60k $
15:29:20 <skalnik1> if we want to measure changes on the level ~1W or ~0.1
15:29:22 <skalnik1> W
15:29:59 <skalnik1> the third way how to measure (scomes) is indirect
15:30:11 <skalnik1> we don't measure watts but CPU usage
15:30:35 <skalnik1> this is built up on systemtap package
15:31:02 <zodbot> Announcement from my owner (stickster): Fedora Board public IRC meeting in ~30 minutes -- http://fedoraproject.org/wiki/Board/IRC for details
15:32:28 <mgahagan> skalnik1, yeah, if we can accurately track cpu & disk usage we are probably 90% of the way there... then you have the display which we can't do that much about since it needs to be bright enough to be usable :)
15:35:51 <skalnik1> mgahagan: right, this was a part of my presentation today - the cpu and graphics consumes aprox. 2/3 power of whole computer
15:36:08 <skalnik1> then important power eater is lcd
15:36:33 <skalnik1> the rest is for hdd, wifi, some usb periphery etc.
15:37:11 <mgahagan> skalnik1, good point about the graphics... do we have something that measures GPU activity.
15:37:26 <skalnik1> currently not
15:38:04 <mjg59> Some CPU use is more expensive than others
15:38:10 <skalnik1> mgahagan:  i'd like to redirect this question to mjg59 :-)
15:38:20 <skalnik1> oh he reacted already :-)
15:38:40 <mgahagan> the biggest power usage on my laptop (at least when idle) turned out to be bluetooth and USB (probably the fingerprint reader)... killing those gives me about an hour of extra battery life if I'm just doing general web browsing.
15:38:44 <mjg59> It's much more power efficient to spend 0.5 seconds active and 0.5 seconds idle than 0.25 seconds active, 0.25 seconds idle, 0.25 seconds active, 0.25 seconds idle
15:38:50 <mgahagan> usbautosuspend seemed to help the most
15:39:08 <mjg59> USB autosuspend should be enabled by default on bluetooth and fingerprint readers in rawhide
15:39:57 <skalnik1> mgahagan: i measured wifi + bluetooth a couple of days ago
15:40:03 <skalnik1> the result is 1.1W
15:40:19 <skalnik1> there is one let say problem
15:40:23 <mgahagan> mjg59, I'm getting ready to try a rawhide live image to see how it does with power consumption... I noticed that rawhide in a KVM guest has much fewer wakeups than F11 did.
15:41:04 <mgahagan> the latest F11 kernel helped some too... CPU can sit in C3 state most of the time now
15:51:32 <skalnik1> well, we are near to end of our time slot. any other remarks, questions?
15:55:37 <skalnik1> so thanks everyone for participating, and see you again next week!
15:55:44 <skalnik1> #endmeeting