[zfs-discuss] Improving read IOPS on RAIDZ

Chris Najmi amir.c.najmi at gmail.com
Thu Apr 28 09:14:21 EDT 2016


Generally ZFS performance scales with number of vdevs, not with number of disks total, or number of disks in a vdev.

I think it is a reasonable assumption that a single vdev raidz of N disks will not perform as well as, for example, an mdadm raid 5 of N disks.

OP, the easiest way to improve read IOPS is to organize your disks into multiple vdevs, and not make pools that consist of single giant vdevs. Next, if you're dealing with largely random IO consider adding more memory and/or an l2arc.

-----Original Message-----
From: "Edward Ned Harvey (zfsonlinux) via zfs-discuss" <zfs-discuss at list.zfsonlinux.org>
Sent: ‎4/‎28/‎2016 9:01 AM
To: "zfs-discuss at list.zfsonlinux.org" <zfs-discuss at list.zfsonlinux.org>; "Hans Henrik Happe" <happe at nbi.ku.dk>
Subject: Re: [zfs-discuss] Improving read IOPS on RAIDZ

> From: zfs-discuss [mailto:zfs-discuss-bounces at list.zfsonlinux.org] On Behalf
> Of Hans Henrik Happe via zfs-discuss
> 
> I had hoped that turning off checksumming would make RAIDZ give similar
> read IOPS as traditional RAID, 

Why would you believe raidz has lower read iops than hardware raid? I believe your presumption is false.
_______________________________________________
zfs-discuss mailing list
zfs-discuss at list.zfsonlinux.org
http://list.zfsonlinux.org/cgi-bin/mailman/listinfo/zfs-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20160428/79e5d4ad/attachment.html>


More information about the zfs-discuss mailing list