Synology SHR calculator

How SHR uses drives of different sizes

Classic RAID treats every drive as the size of the smallest one. Put two 4 TB and two 8 TB drives into RAID 5 and you get 12 TB: the extra 4 TB on each large drive is wasted. SHR gives you 16 TB from the same drives.

It does this by cutting the drives into layers. The first layer is as tall as the smallest drive and spans all of them. The next layer covers what is left on the larger drives, and so on. Each layer gets its own RAID: RAID 5 where it spans three or more drives, a mirror where it spans two. In SHR-2 the layers use RAID 6, or a three-way mirror where only three drives reach that high. All layers are then joined into one volume.

For 2, 4, 4 and 8 TB drives in SHR, the layers are: 2 TB across four drives in RAID 5 (6 TB), 2 TB across three drives in RAID 5 (4 TB), and 4 TB on the 8 TB drive alone, which has no partner and stays unused. Total: 10 TB.

That gives a simple rule. SHR: all drives added together, minus the largest. SHR-2: all drives minus the two largest. If the largest drive is bigger than every other drive, the difference is unused until you add a second drive of that size.

SHR or classic RAID 5 and RAID 6

With drives of equal size, SHR is exactly RAID 5 and SHR-2 is exactly RAID 6: same space, same protection, same speed. The difference only appears with mixed sizes, and when you upgrade later.

Under the hood, SHR is ordinary Linux software RAID (mdadm) with the layers joined by LVM. That is good news if the NAS itself dies: the drives can be read on a Linux computer, and Synology documents how to do it.

Classic RAID has one advantage: it is available on every Synology model. The larger business models (the XS series and many rack-mount units) do not offer SHR by default.

SHR or SHR-2

SHR survives one failed drive. While the replacement rebuilds, there is no protection left, and with large drives a rebuild takes a day or more, reading every other drive end to end. A second failure, or a read error on another drive, can cost data in that window.

SHR-2 survives two failed drives, at the cost of a second drive's worth of space, and needs at least four drives. With five or more drives, or drives of 12 TB and more, it is the safer choice. The drive failure probability calculator shows how likely a failure is in your array, and the rebuild time calculator how long the exposed window lasts.

Upgrading to larger drives

You can grow an SHR pool by replacing drives one at a time with larger ones, waiting for each rebuild to finish. The table shows the catch: the first larger drive adds nothing, because it has no partner. In SHR the space grows from the second large drive on, in SHR-2 from the third.

Adding a drive to an empty bay is quicker: it joins the existing layers straight away. The table under the calculator shows what each drive size would add to your current setup.

Why DSM shows less than the calculator

The calculator counts in decimal terabytes, the way drives are sold. DSM shows capacity in binary units, which makes every figure about 9% smaller (the calculator shows the TiB figure under the main number). DSM also keeps a small system partition on every drive for itself, and the file system needs some space for its own records, so expect the free space on a new volume to be a little lower still.

Questions