Date: Mon, 22 Jun 2015 08:17:35 -0500 (CDT) From: Bob Friesenhahn <bfriesen@simple.dallas.tx.us> To: Quartz <quartz@sneakertech.com> Cc: FreeBSD FS <freebsd-fs@freebsd.org> Subject: Re: ZFS raid write performance? Message-ID: <alpine.GSO.2.01.1506220802100.4186@freddy.simplesystems.org> In-Reply-To: <5587C3FF.9070407@sneakertech.com> References: <5587C3FF.9070407@sneakertech.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 22 Jun 2015, Quartz wrote: > What's sequential write performance like these days for ZFS raidzX? Someone > suggested to me that I set up a single not-raid disk to act as a fast > 'landing pad' for receiving files, then move them to the pool later in the > background. Is that actually necessary? (Assume generic sata drives, > 250mb-4gb sized files, and transfers are across a LAN using single unbonded > GigE). The primary determinant of write performance is if the writes are synchronous or not, With synchronous writes, the data is comitted to non-volatile storage before responding to the requestor. With asyncronous writes, the data only needs to be written into RAM before responding to the requestor. Writes over NFS 3 are synchronous. Writes over CIFS/Samba are likely not. For good performance with synchronous writes, some sort of non-volatile write cache (e.g. dedicated zfs intent log "slog", controller NVRAM) is advised. Use multiple sets of mirrors for maximum write performance with multiple clients. Even 10 years old hardware should be able to keep up with gigabit Ethernet rates (< 100MB/s) given a reasonable disk subsystem. 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?alpine.GSO.2.01.1506220802100.4186>