Date: Fri, 1 Jun 2012 11:04:39 +0200 (CEST) From: Wojciech Puchar <wojtek@wojtek.tensor.gdynia.pl> To: Daniel Staal <DStaal@usa.net> Cc: Dennis Glatting <dg@pki2.com>, freebsd-questions@freebsd.org Subject: Re: Anyone using freebsd ZFS for large storage servers? Message-ID: <alpine.BSF.2.00.1206011101190.2497@wojtek.tensor.gdynia.pl> In-Reply-To: <70D99F583BF63C05BB39CCAF@mac-pro.magehandbook.com> References: <CACxnZKM__Lt9LMabyUC_HOCg2zsMT=3bpqwVrGj16py1A=qffg@mail.gmail.com> <alpine.BSF.2.00.1205311102290.48219@btw.pki2.com> <70D99F583BF63C05BB39CCAF@mac-pro.magehandbook.com>
next in thread | previous in thread | raw e-mail | index | archive | help
> I'm not using as huge a dataset, but I was seeing this behavior as well when > I first set my box up. What was happening was that ZFS was caching *lots* of > writes, and then would dump them all to disk at once, during which time the > computer was completely occupied with the disk I/O. > > The solution (suggested from <http://wiki.freebsd.org/ZFSTuningGuide>) for me > was: > vfs.zfs.txg.timeout="5" both problem, and solution is very close to linux style ext2/3/4 and it's behaviour. And one of the main reason to moving out from this s..t to FreeBSD. (the other was networking) UFS writes out complete MAXBSIZE sized chunks quickly. all of that behaviour or linux (and probably ZFS) are because it often gives better result in benchmark, and people love synthetic benchmarks.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.00.1206011101190.2497>