Choosing the amount of SSD storage capacity - FSx for ONTAP

Choosing the amount of SSD storage capacity

When choosing amount of SSD storage capacity for your FSx for ONTAP file system, you need to keep in mind the following items that impact the amount of SSD storage available for storing your data:

  • Storage capacity reserved for the NetApp ONTAP software overhead.

  • File metadata

  • Recently written data

  • Files that you intend to store on SSD storage, whether it's data that hasn't hit its cooling period, or data that you recently read that was retrieved back to SSD.

How SSD storage is used

Your file system's SSD storage is used for a combination of NetApp ONTAP software (overhead), file metadata, and any data stored on your SSD tier.

NetApp ONTAP software overhead

Like other NetApp ONTAP file systems, 16% of a file system's SSD storage capacity is reserved for ONTAP overhead, as follows:

  • 11% is reserved for NetApp ONTAP software.

  • 5% is reserved for aggregate snapshots, which are required to synchronize data between both of a file system's file servers.

The SSD capacity reserved by ONTAP is not available for storing your files.

File metadata

File metadata typically consumes 3-7% of the storage capacity that is consumed by the files. This percentage depends on the average file size (a smaller average file size requires more metadata), and the amount of storage efficiency savings achieved on your files. Note that file metadata doesn't benefit from storage efficiency savings. You can use the following guidelines for estimating the amount of SSD storage used for metadata on your file system.

Average file size Size of metadata as a percentage of file data

4 KB

7%

8 KB

3.5%

32 KB or greater

1-3%

When sizing the amount of SSD storage capacity you need for the metadata of files you plan to store on the capacity pool tier, we recommend using a conservative ratio of 1 GiB of SSD storage for every 10 GiB of data you plan to store on the capacity pool tier.

File data stored on your SSD tier

In addition to your active data set and all file metadata, all data written to your file system is initially written to the SSD tier before being tiered-off to capacity pool storage. This is true regardless of the volume's tiering policy, with the exception of transferring data using SnapMirror to a volume configured with an All data tiering policy.

Random reads from the capacity pool tier are cached in the SSD tier, as long as the SSD tier is under 90% utilization. For more information, see Volume data tiering.

Recommended SSD capacity utilization

We recommend that you do not exceed 80% utilization of your SSD storage tier in an ongoing basis. This recommendation is consistent with NetApp's recommendation for ONTAP. Because your file system’s SSD tier is also used for staging writes to, and for random reads from, the capacity pool tier, any sudden changes in access patterns can quickly cause the utilization of your SSD tier to increase.

At 90% SSD utilization, data read from the capacity pool tier is no longer cached on the SSD tier so that the remaining SSD capacity is preserved for any new data that is written to the file system. This causes repeat reads of the same data from the capacity pool tier to be read from capacity pool storage instead of being cached and read from the SSD tier, which can impact the throughput capacity your file system.

All tiering functionality stops when the SSD tier is at or above 98% utilization. For more information, see Tiering thresholds.

SSD sizing example

Assume you want to store 100 TiB of data for an application where 80% of the data is infrequently accessed. In this scenario, 80% (80 TB) of your data is automatically tiered to the capacity pool tier and the remaining 20% (20 TB) remains in SSD storage. Based on the typical storage efficiency savings of 65% for general-purpose file sharing workloads, that equates to 7 TiB of data. To maintain an 80% SSD utilization rate, you need 8.75 TiB of SSD storage capacity for the 20 TiB of actively-accessed data. The amount of SSD storage that you provision also needs to account for the ONTAP software storage overhead of 16%, as shown in the following calculation.

ssdNeeded = ssdProvisioned * (1 - 0.16) 8.75 TiB / 0.84 = ssdProvisioned 10.42 TiB = ssdProvisioned

So in this example, you need to provision at least 10.42 TiB of SSD storage. You will also use 28 TiB of capacity pool storage for the remaining 80 TiB of infrequently accessed data.