Date: Wed, 12 Jun 2013 07:03:58 -0500 From: "Mark Felder" <feld@feld.me> To: "Jeremy Chadwick" <jdc@koitsu.org> Cc: freebsd-fs@freebsd.org Subject: Re: An order of magnitude higher IOPS needed with ZFS than UFS Message-ID: <op.wykegwri34t2sn@markf.office.supranet.net> In-Reply-To: <20130612114937.GA13688@icarus.home.lan> References: <51B79023.5020109@fsn.hu> <op.wykdduw834t2sn@markf.office.supranet.net> <20130612114937.GA13688@icarus.home.lan>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 12 Jun 2013 06:49:37 -0500, Jeremy Chadwick <jdc@koitsu.org> wrote: > > Mark, do you have any references for this? I'd love to learn/read more > about this engineering/design aspect (I won't say flaw, I'll just say > aspect) to ZFS, as it's the first I've heard of it. Firsthand experience on a couple servers, and some old Sun docs that I can't find anymore since Oracle broke the links. If you start googling for "ZFS performance 80%" you should come across similar reports. The recommendation was always that when you hit about 80% you need to add a new vdev or you'll be in serious trouble. I'd always believed that it has to do with the way the ZFS COW algorithm works. If my suspicion is correct I'd guess it probably stalls trying to find an ideal place to write -- maybe some cost calculation? I'm reaching for straws now because I don't know anything about the code itself. I'd love to hear from people who have actually touched the code and can give a more definitive answer because this does border on "urban legend" territory, but I've read it and experienced it a few times so I'm just passing it on.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?op.wykegwri34t2sn>