NetApp AFF and Advanced Drive Partition v2, part 3

In previous posts I wrote about AFF and ADPv2 with 12 SSD disks, you can find these two entries here and here

With AFF product family ADPv2 supports slicing up to 48 SSD disks. Let’s investigate larger setups, starting with 24 SSD disk setups.

Example of 24×3.8TB SSD setup with ADPv1. With 24 disks there are enough slices to make two data aggregates even with ADPv1. However ADPv1 uses more parity and spare slices than ADPv2, both aggregates have their own parity slices and spare slices .

With 24×3,8TB and ADPv1 you will get about 61,27 TB of usable capacity or about 67,18% usable capacity out of raw capacity.

The same example with 24×3,8TB SSD and ADPv2. With ADPv2 data partitions have shared parity and spare drives, resulting in higher usable capacity than with ADPv1.

With 24×3,8TB and ADPv2 you will get about 71,48 TB of usable capacity or about 78,37% usable capacity out of raw capacity.

With ADPv2 and 24 drives you will get more usable capacity than with ADPv1. How much more? ((71,48 – 61,27) / 61,27 ) * 100 = 16,67 %

What about even larger setups? Normally disks in NetApp environments are sold at half-a-shelf increments, so the next logical step would be 36xSSD setup. Let’s use 15,3 TB SSD drives this time.

Note: 15,3 TB SSD drives are only supported with high-end AFF8080EX controller (and Ontap 9), I guess that the lesser controller models (AFF8040 & AFF8060) don’t have enough oomph to fully utilize the massive capacity that 15,3 TB SSD drives provide. Additionally these drives require different shelf, a new DS224C with 12 Gb/s SAS connectivity.

Example of 36×15,3TB SSD drives with ADPv1:

With ADPv1 and 36×15,3TB SSD drives you will get about 413,88 TB of usable capacity or 75,14% usable capacity out of raw marketing capacity. Wow, that is quite a lot of space in just ten rack units (6U for the controller + 2 x 2U for the two shelves). In 2001 I was working in Singapore and we had state-of-the-art EMC Symmetrix box, which was about the size of a delivery van. Usable space for that one was about 2TB, built out of tiny 32GB disks…

Example of 36×15,3TB drives with ADPv2:

With ADPv2 and 36×15,3TB SSD drives you will get the same usable capacity as with the same setup with ADPv1. I.e. NO capacity benefit in using ADPv2. Why is that? I comes down to maximum spindle/drive/slice count per raid group. With RAID-DP protection level the maximum size of a raid group with SSD drives is 28 drives (or slices). With ADPv2 and 36 disks you have 34 slices to use per data aggregate. Since these won’t fit into a one raid group, you will have to use two raid groups per data aggregate. By using two raid groups, you will increase number of parity and spare slices and decrease number slices for data.

The next  logical step in disk count would  be 48 disks. This will have the same end result as  setup with 36 disks, there is no difference in usable capacity between ADPv1 and  ADPv2, because of multiple raid group requirement.

Conclusion: Using ADPv2 with 12 and 24 SSD disk setups will increase usable capacity. Larger setups up to 48 SSD disks will use the same slicing scheme with ADPv2, but since you need more raid groups with larger setups, there is no capacity advantage.

In part 4 I will investigate  AFF / SSD setups with about the same usable capacity, but using different SSD sizes to achieve the usable capacity.