Date: Thu, 30 Mar 2000 17:43:07 -0800 (PST) From: Matthew Dillon <dillon@apollo.backplane.com> To: Warner Losh <imp@village.org> Cc: "David O'Brien" <obrien@FreeBSD.ORG>, freebsd-current@FreeBSD.ORG Subject: Re: SMP buildworld times / performance tests Message-ID: <200003310143.RAA71466@apollo.backplane.com> References: <20000330180402.E10480@futuresouth.com> <Pine.BSF.4.10.10003301158360.14426-100000@backup.af.speednet.com.au> <200003300604.WAA68031@apollo.backplane.com> <20000330174453.D10480@futuresouth.com> <20000330160043.A57508@dragon.nuxi.com> <200003310058.RAA30753@harmony.village.org>
next in thread | previous in thread | raw e-mail | index | archive | help
:In message <20000330180402.E10480@futuresouth.com> "Matthew D. Fuller" writes:
:: The question at task is, is buildworld one of them? I don't think that
:: situation comes up a lot in buildworld, but I'm not exactly an authority
:: on it...
:
:About 6 months ago, softupdates made things about 5% faster than async
:for makeworld on my PPro 200 + 196M of memory.
:
:Warner
Softupdates is basically going to beat async on just about everything,
and beat it *badly* for a bunch of things like /tmp operation.
There are a few minor issues with the write-behind code (e.g. that
thread on DBM ops and certain contrived random I/O tests being slow),
which async is much faster on, but that's just a quirk in the clustering
code. I'm working on a patch set which extends the sequential read
heuristic to also handle writes. The patch is currently under review
and can hopefully be committed to 4.x and 5.x soon (or something like
it).
Async should not be used unless you really like restoring crashed
filesystems from tape :-). Oh, and perhaps when one is doing an
initial OS install from CDRom :-). Async itself will not cause a crash,
but if your machine crashes in the middle of a bunch of async writes
you might end up with an unrecoverable filesystem.
Also, async can be awefully hard on the VM system if you are doing a
lot of writing - there's a reason why we have write-behind code, even
if it has a few bogus cases.
-Matt
Matthew Dillon
<dillon@backplane.com>
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200003310143.RAA71466>