Date: Thu, 10 Oct 2002 10:35:31 -0700 (PDT) From: Scott Hess <scott@avantgo.com> To: freebsd-hackers@FreeBSD.ORG Subject: Re: Mount option "nomtime"? Message-ID: <Pine.LNX.4.44.0210101011290.1689-100000@river.avantgo.com> In-Reply-To: <200210100600.g9A60xiX034084@lurza.secnetix.de>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 10 Oct 2002, Oliver Fromme wrote: > Wolfieee <mjoyner@hq.dyns.cx> wrote: > > what about vnconfiging some files on the netapp thing, this way the > > *bsd can see the vnconfig files as local raw disk partitions? > > That wouldn't make a difference, because the NetApp Filer would still > schedule mtime updates for evey write access to the file that has been > vnconfiged on the client. Probably talking through my hat, here, but doesn't NetApp use a LFS (log-structured file system) variant? It's been a couple years since we got ours, but at the time I reviewed it for use as a central database filesystem. I'm pretty sure it used an LFS, partially because of the sweet RAID-write semantics - basically, you never have to read-modify-write to deal with the parity update, you simply generate an entire chunk worth of data in NVRAM, and write it all at once, with parity. In any case, assuming that I got the analysis right, that means the mtime update should be essentially free, since it's already going to be writing a new inode for the file to capture any new blocks written. http://www.netapp.com/tech_library/3001.html#I4, also note the Ousterhout reference in the bibliography. An even stronger indicator is at http://www.netapp.com/tech_library/3002.html#I35. Later, scott To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.4.44.0210101011290.1689-100000>