Date: Thu, 29 Mar 2012 16:11:54 -0400 (EDT) From: Rick Macklem <rmacklem@uoguelph.ca> To: Bob Friesenhahn <bfriesen@simple.dallas.tx.us> Cc: freebsd-fs@freebsd.org Subject: Re: NFSv3, ZFS, 10GE performance Message-ID: <2042316157.1946865.1333051914085.JavaMail.root@erie.cs.uoguelph.ca> In-Reply-To: <alpine.GSO.2.01.1203281650280.25475@freddy.simplesystems.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Bob Friesenhahn wrote: > On Wed, 28 Mar 2012, Rick Macklem wrote: > >> > >> Hopefully, readahead doesn't kill performance for smaller files.. > >> :-) > >> > > Well, readaheads only happen if the file is large enough for the > > readahead > > to be before EOF. As such, they just won't happen for small files. > > The real problem is for applications which do a seek and a new read > prior to consuming all the data which is being read ahead on its > behalf. This results in read amplification and increased latency for > the next seek+read. > Yes, random reads on a large file is a problem, as you noted before. (I was just clarifying the "small file" case and it's good that you are clarifying the "large file, non-sequential" case.) As jhb@ noted, disabling read-ahead may be a good idea for the POSIX hint. Thanks, rick > Bob > -- > Bob Friesenhahn > bfriesen@simple.dallas.tx.us, > http://www.simplesystems.org/users/bfriesen/ > GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?2042316157.1946865.1333051914085.JavaMail.root>