Date: Sun, 20 Jan 2008 13:17:31 -0600 From: Stephen Montgomery-Smith <stephen@math.missouri.edu> To: Jeremy Chadwick <koitsu@FreeBSD.org> Cc: freebsd-stable <freebsd-stable@FreeBSD.org>, "Jason C. Wells" <jcw@highperformance.net>, TooMany Secrets <toomany@toomany.net> Subject: Re: Scheduler in Various Docs Message-ID: <47939E4B.5050706@math.missouri.edu> In-Reply-To: <20080120191316.GA13382@eos.sc1.parodius.com> References: <479388C0.50507@highperformance.net> <47938F21.6020308@math.missouri.edu> <e8b5dfd50801201040s755d7dc4p962095d63b57503a@mail.gmail.com> <20080120191316.GA13382@eos.sc1.parodius.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Jeremy Chadwick wrote: > On Sun, Jan 20, 2008 at 07:40:07PM +0100, TooMany Secrets wrote: >> On 1/20/08, Stephen Montgomery-Smith <stephen@math.missouri.edu> wrote: >>> Jason C. Wells wrote: >>>> The comments regarding SCHED_ULE and SCHED_4BSD are inconsistent with >>>> information found in the email archives. LINT says ULE is experimental. >>>> The handbook doesn't mention ULE at all. The archives say ULE is the >>>> new recommended scheduler. >>>> >>>> If ULE is in fact the current recommendation, then a few docs need to be >>>> updated. >>> To add to Jason's point - why does GENERIC still default to SCHED_4BSD? >>> Are there plans to change this before 7.0 is truly released? >> Excuse me for my bad english... >> >> This question was mentioned two or three months ago. The answer was >> that in 7.1, after the ULE will be tested in 7.0, it will be the >> defacto scheduler in FreeBSD. First, the scheduler need the best >> benchark in the world; a few thousand users testing in real-life >> situations on a daily basis. > > This is correct. There was a very large discussion on freebsd-current > (which would've been discussing 7.x at that point) about what scheduler > should be the default for RELENG_7 (4BSD or the "new" ULE (a.k.a. > SMP2)). It was ""voted"" (note the quotes) that SCHED_4BSD should > remain the default until 7.1 was released, since if there turned out > to be a gigantic bug in the new scheduler, we wouldn't want people to > get bit by it (thus harming the stability reputation of -RELEASE and > -STABLE). The 4BSD scheduler is still considered stable and has a > track record to prove it. > > In a way, SCHED_ULE on 7.x is still considered "experimental" in the > sense that it needs lots of people testing it. So far all the results > have been positive (unlike SCHED_ULE on 6.x and 5.x, which were very > broken -- hence the rewrite!). > > If the OP wants to read the thread/discussion (it's long), I can dig up > a URL to it in the archives. Thanks. You both answered my question admirably.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?47939E4B.5050706>