Date: Thu, 27 May 2004 16:07:45 +0200 From: Bernd Walter <ticso@cicely12.cicely.de> To: Ivan Voras <ivoras@fer.hr> Cc: current@freebsd.org Subject: Re: Softupdates a mount option? Message-ID: <20040527140744.GW63479@cicely12.cicely.de> In-Reply-To: <40B5E66F.7000507@fer.hr> References: <40B4ECC8.50808@fer.hr> <20040526202849.GA37162@freebie.xs4all.nl> <40B519DA.7000708@fer.hr> <20040527120819.B8434@gamplex.bde.org> <40B5DE26.4040901@fer.hr> <20040527124512.GV63479@cicely12.cicely.de> <40B5E66F.7000507@fer.hr>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, May 27, 2004 at 03:00:31PM +0200, Ivan Voras wrote: > Bernd Walter wrote: > > >>>It makes sense to never enable soft updates on a memory drive, since soft > >>>updates uses extra CPU cycles to try to speed up i/o to real drives (and > >> > >>Then maybe the default should be changed? > >> > >>From 'man mdmfs': > >> By default, mdmfs creates a swap-based (MD_SWAP) disk with > >> soft-updates > >> enabled and mounts it on mount-point. > > > > > >swap != ram > >SU makes perfectly sense for swap backed md drives. > > I always thought the "swap backed" meant the memory is allocated from the > same pool as for userland applications, e.g. they only get swapped out if > memory is scarce. Is this wrong? You are right, but md(4) doesn't know about the filesystem and therefor can't know which blocks have content to keep and which are unused. SU now allows files that are deleted quite fast to never touch the block device and md never need to write those blocks into swap storage as they never got dirty. The memory can be reclaimed without the need for a physical write. The story is different with type malloc where there is not physical device and we only have ram. -- B.Walter BWCT http://www.bwct.de bernd@bwct.de info@bwct.de
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040527140744.GW63479>