Optimizing Performance of SSDs and Advanced Format Drives - Revision history

Malcolm at 05:53, 31 August 2017

2017-08-31T05:53:22Z

← Older revision		Revision as of 22:53, 30 August 2017
Line 29:		Line 29:

	[[Category:ZFS OSD Tuning]]		[[Category:ZFS OSD Tuning]]
			[[Category:Lustre Systems Administration]]

Malcolm at 03:29, 15 August 2017

2017-08-15T03:29:12Z

← Older revision		Revision as of 20:29, 14 August 2017
Line 19:		Line 19:
	<pre style="overflow=x:auto;">		<pre style="overflow=x:auto;">
	$ sudo zpool add -o ashift=12 tank mirror sdc sdd		$ sudo zpool add -o ashift=12 tank mirror sdc sdd
			</pre>

			'''Note:''' Do not set the <code>ashift</code> property to a value that will assign a sector size larger than the device physical block size. The logical and physical block sizes can be read from sysfs:

			<pre style="overflow-x:auto;">
			cat /sys/class/block/<dev>/queue/logical_block_size
			cat /sys/class/block/<dev>/queue/physical_block_size
	</pre>		</pre>

	[[Category:ZFS OSD Tuning]]		[[Category:ZFS OSD Tuning]]

Malcolm at 03:25, 15 August 2017

2017-08-15T03:25:26Z

← Older revision		Revision as of 20:25, 14 August 2017
Line 5:		Line 5:
	The default behavior of ZFS is to automatically detect the sector size of the drive. However, when attempting to detect the sector size of an AF drive, ZFS will not be able to detect the native sector size, and will instead trust the drive when it reports the emulated size. This can result in poorly aligned disk access that degrades performance of the pool.		The default behavior of ZFS is to automatically detect the sector size of the drive. However, when attempting to detect the sector size of an AF drive, ZFS will not be able to detect the native sector size, and will instead trust the drive when it reports the emulated size. This can result in poorly aligned disk access that degrades performance of the pool.

	Therefore, the ability to set the <code>ashift</code> property has been added to the <code>zpool</code> command. This allows users to explicitly assign the sector size when devices are first added to a pool (typically at pool creation time or when adding a <code<vdev</code> to the pool). The <code>ashift</code> values range from 9 to 16, with the default value 0 meaning that ZFS should auto-detect the sector size.		Therefore, the ability to set the <code>ashift</code> property has been added to the <code>zpool</code> command. This allows users to explicitly assign the sector size when devices are first added to a pool (typically at pool creation time or when adding a <code>vdev</code> to the pool). The <code>ashift</code> values range from 9 to 16, with the default value 0 meaning that ZFS should auto-detect the sector size.

	This issue is rare on hard disk drives, and so it might not be necessary to alter the <code>ashift</code> value. However, there are reports that SSDs can benefit from setting the <code>ashift</code> property explicitly to match the 4096 byte sector size.		This issue is rare on hard disk drives, and so it might not be necessary to alter the <code>ashift</code> value. However, there are reports that SSDs can benefit from setting the <code>ashift</code> property explicitly to match the 4096 byte sector size.

Malcolm: Created page with "The following information is taken from a FAQ entry on the ZFS on Linux project web site, reference: https://github.com/zfsonlinux/zfs/wiki/faq. Advanced Format (AF) is a dis..."

2017-08-14T08:28:17Z

Created page with "The following information is taken from a FAQ entry on the ZFS on Linux project web site, reference: https://github.com/zfsonlinux/zfs/wiki/faq. Advanced Format (AF) is a dis..."

New page

The following information is taken from a FAQ entry on the ZFS on Linux project web site, reference: https://github.com/zfsonlinux/zfs/wiki/faq.

Advanced Format (AF) is a disk format that natively uses a sector size of 4,096 bytes instead of 512 bytes. To maintain compatibility with legacy systems, AF disks emulate a sector size of 512 bytes.

The default behavior of ZFS is to automatically detect the sector size of the drive. However, when attempting to detect the sector size of an AF drive, ZFS will not be able to detect the native sector size, and will instead trust the drive when it reports the emulated size. This can result in poorly aligned disk access that degrades performance of the pool.

Therefore, the ability to set the <code>ashift</code> property has been added to the <code>zpool</code> command. This allows users to explicitly assign the sector size when devices are first added to a pool (typically at pool creation time or when adding a <code<vdev</code> to the pool). The <code>ashift</code> values range from 9 to 16, with the default value 0 meaning that ZFS should auto-detect the sector size.

This issue is rare on hard disk drives, and so it might not be necessary to alter the <code>ashift</code> value. However, there are reports that SSDs can benefit from setting the <code>ashift</code> property explicitly to match the 4096 byte sector size.

The value of <code>ashift</code> is actually a bit shift value, so the <code>ashift</code> value for 512 bytes is 9 (2<sup>9</sup> = 512) while the <code>ashift</code> value for 4,096 bytes is 12 (2<sup>12</sup> = 4,096). To force the pool to use 4,096 byte sectors at pool creation time:

<pre style="overflow=x:auto;">
$ sudo zpool create -o ashift=12 tank mirror sda sdb
</pre>

To force the pool to use 4,096 byte sectors when adding a vdev to a pool:

<pre style="overflow=x:auto;">
$ sudo zpool add -o ashift=12 tank mirror sdc sdd
</pre>

[[Category:ZFS OSD Tuning]]