Date: Sun, 30 Jul 2006 14:11:40 -0700 (PDT) From: "R. B. Riddick" <arne_woerner@yahoo.com> To: David Godsey <freebsd@godseyfamily.com>, freebsd-realtime@freebsd.org Subject: Re: Information Sites Message-ID: <20060730211140.24074.qmail@web30310.mail.mud.yahoo.com> In-Reply-To: <3945.71.213.169.51.1154290426.squirrel@godseyfamily.com>
next in thread | previous in thread | raw e-mail | index | archive | help
--- David Godsey <freebsd@godseyfamily.com> wrote: > So I have been combing the internet for information on realtime FreeBSD, > but the information is limited. So there must be some more information > out there, but I do not know where to look. Can someone direct me please? > Hmm... I do not have practical experience in real time operations (just my mostly theoretical studies of task scheduling and control engineering), so I can only hope my contribution is not too useless/stupid... :-) I say, have u read the man page rtprio(2) and polling(4)? I would guess, that u might want to set kern.clockrate to something high (1000..2000) (e. g. via /boot/loader.conf's kern.hz). Furthermore I would recommend, that u do some tests/research/experiments in ur specific environment (remember: real time constraints can be very different (e. g. a simple electronic calculator is ok, when u just want to compute the taxes in a shop or in a restaurant, but when u want to compute the course of Cpt. Picard's Enterprise (StarTrek) u would need something faster (or it might get struck by some kind of lightning or get stuck in a sun))). In this case here it is the question, if u can tolerate fluctuations in scheduling accuracy (response time) due to voluntary/unpredictable/dynamically-scheduled kernel activity (I think polling(4) could just help (a little) with network devices, so that hard disc activity can still hinder real time tasks). >From the point of view of digital control engineering operating systems like *BSD are no real real-time operating systems... Real real-time operating systems are obviously able to guarantee a very high scheduling accuracy... I just did this litte experiment with (avg. about 1.3Mbit/sec, peak about 16Mbit/sec) disc traffic at the same time and a clock rate of 1000Hz: > rtprio rtprio: normal priority > repeat 10 time sleep .2 0.007u 0.015s 0:00.22 4.5% 12+396k 0+0io 0pf+0w 0.000u 0.030s 0:00.23 13.0% 182+1332k 0+0io 0pf+0w 0.000u 0.032s 0:00.23 13.0% 181+1288k 0+0io 0pf+0w 0.010u 0.021s 0:00.23 13.0% 276+2064k 0+0io 0pf+0w 0.006u 0.026s 0:00.24 8.3% 144+1230k 0+0io 0pf+0w 0.000u 0.029s 0:00.23 8.6% 136+966k 0+0io 0pf+0w 0.000u 0.032s 0:00.24 12.5% 185+1420k 0+0io 0pf+0w 0.021u 0.010s 0:00.24 12.5% 182+1332k 0+0io 0pf+0w 0.006u 0.025s 0:00.28 7.1% 278+2130k 0+0io 0pf+0w 0.008u 0.026s 0:00.24 8.3% 276+2064k 0+0io 0pf+0w neo# rtprio 30 -$$ neo# rtprio rtprio: realtime priority 30 neo# repeat 10 time sleep .3 0.000u 0.025s 0:00.33 6.0% 138+1018k 0+0io 0pf+0w 0.000u 0.028s 0:00.33 6.0% 276+2036k 0+0io 0pf+0w 0.000u 0.027s 0:00.33 6.0% 140+1084k 0+0io 0pf+0w 0.000u 0.028s 0:00.33 6.0% 138+1018k 0+0io 0pf+0w 0.006u 0.024s 0:00.32 6.2% 412+2988k 0+0io 0pf+0w 0.000u 0.026s 0:00.32 6.2% 138+1018k 0+0io 0pf+0w 0.006u 0.020s 0:00.47 4.2% 274+1970k 0+0io 0pf+0w 0.006u 0.019s 0:00.32 3.1% 552+4072k 0+0io 0pf+0w 0.000u 0.013s 0:00.31 3.2% 276+2036k 0+0io 0pf+0w 0.000u 0.014s 0:00.31 3.2% 276+2036k 0+0io 0pf+0w -Arne __________________________________________________ Do You Yahoo!? Tired of spam? Yahoo! Mail has the best spam protection around http://mail.yahoo.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060730211140.24074.qmail>