[prev in list] [next in list] [prev in thread] [next in thread] 

List:       busybox
Subject:    Re: tickless cron
From:       Denis Vlasenko <vda.linux () googlemail ! com>
Date:       2006-09-26 12:39:13
Message-ID: 200609261439.13657.vda.linux () googlemail ! com
[Download RAW message or body]

On Monday 25 September 2006 21:41, Rob Landley wrote:
> On Sunday 24 September 2006 5:52 pm, Denis Vlasenko wrote:
> > Periodic check every minute really isn't a load worth optimizing out.
> 
> On a cell phone it sure is.  Their standby power budget is about 5 miliwatts.  
> A device with a week of standby and three hours of talk time really doesn't 
> want to power up to "doing something" mode unless it gets the wake-on-lan 
> call-coming-in bat-signal, or unless you open the display to get its 
> attention.  Waking up once a minute even for a few miliseconds, round the 
> clock including 4 am, will chop a couple _days_ off the standby time.

Run time per day:
10 milliseconds * 1440 minutes = 14400 milliseconds = 14.4 seconds.
In a week, that amounts to one minute of run time.

OTOH, powerup/sleep cycles may drain extra power by themself, so
it maybe can drain significantly more than that.

OTOH#2, who says that we can't check every 15 minutes?

I think than cron have to act semi-sensibly when someone sets system time:
if we have one task which should be run at 06:00, we shouldn't sleep
for 24 hours in between in one giant nanosleep(). If someone will
adjust system time +5 minutes, we will miss next run - even if it was
still many hours ahead at a moment of system time adjustment!

Is there a way to have event/signal from kernel on system time changes?
--
vda
_______________________________________________
busybox mailing list
busybox@busybox.net
http://busybox.net/cgi-bin/mailman/listinfo/busybox
[prev in list] [next in list] [prev in thread] [next in thread]