Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 6 Sep 2011 09:45:40 -0700
From:      David Brodbeck <brodbd@uw.edu>
To:        freebsd-fs@freebsd.org
Subject:   Re: ZFSv28+NFSv4 poor file creation performance, "sync=disabled" has no effect
Message-ID:  <CAHHaOuZ9ObnjMz48=v0sn1G1A4XHYqpE9v=o173B%2BX9W0fpuiA@mail.gmail.com>
In-Reply-To: <20110904012855.GA79580@icarus.home.lan>
References:  <CAHHaOuY=BEMrhYuzXtD5AtXG7niLXEO1yhO5P4EimcsLuTrLXw@mail.gmail.com> <14220705.747900.1315013770239.JavaMail.root@erie.cs.uoguelph.ca> <CAHHaOua8i8ZvRpbtev0knFJCW0m0i9PSD_w3U0J9FgU44oAW=A@mail.gmail.com> <CAHHaOuaK9Cedq5LFskQMjAH2qu5ukGRNK1DfHEYmZ_hRUbqdYg@mail.gmail.com> <20110904012855.GA79580@icarus.home.lan>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Sep 3, 2011 at 6:28 PM, Jeremy Chadwick <freebsd@jdc.parodius.com>wrote:

> > Also, most other filesystems make no guarantees that data has been
> > flushed to the platters, only that it's been sent to the disk
> > hardware, so it really isn't a big increase in risk compared to our
> > pre-ZFS setup.
>
> Can you explain where you got this impression from?  Please provide
> actual references/validation/proof for non-network filesystems.
>

My understanding is most filesystems only flush to the hardware disk cache,
they don't force the disks to write to the platters like ZFS does, and
that's why ZFS's NFS performance is slow compared to, say, UFS. It could be
I'm mistaken.

What we (the community) need clarification regarding is whether or not
> OpenSolaris and/or Illumos still provides the zil_disable tunable on
> those OSes; if they do, FreeBSD should provide the same (which means
> removable of zil_disable on FreeBSD is effectively a regression).  If
> said OSes do not provide it, then FreeBSD should not provide it.  If
> said OSes moved to using the sync parameter where it works, yet it does
> not work on FreeBSD, then that's a bug and a PR should be filed +
> discussion induced.
>

My understanding is the upstream filesystems removed that feature and added
the sync parameter.

Testing against OpenIndiana to see whether sync makes a difference there is
on my to-do list.  I may get to it today; if so I'll let you know what I
find.


> The risks/complexities and performance concerns with SSDs when used as
> part of a ZFS filesystem on FreeBSD greatly differ from that of using an
> SSD as a dedicated intent log (ZIL) device.
>

Could you elaborate?  I'm hearing a lot of stories of massive slowdowns
after SSDs have been in use for a while, which doesn't sound promising if
I'm going to add one to improve performance.  Also my understanding is
losing the ZIL means irretrievably losing the whole storage pool, so it
creates another point of failure.


>
> What exactly do you need ZFS for given your requirements?  Is it that
> you want a simple-to-use LVM-like filesystem that offers a vast
> in-memory cache but without the data integrity bits?
>

It's mostly the easy filesystem management I'm after.  The problem with
LVM-like solutions is you don't generally get a shared free space pool and
expanding or shrinking a filesystem is a multi-step manual process.


-- 
David Brodbeck
System Administrator, Linguistics
University of Washington



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAHHaOuZ9ObnjMz48=v0sn1G1A4XHYqpE9v=o173B%2BX9W0fpuiA>