Synology NAS DSM SSD Caching

[Enquiries]

I have a Synology DS925+ that I am setting up for a small business (between 10 and 20 users). It will be used primarily for two purposes:
1. A local network fileshare (mostly documents and small files, but also some larger video files)
2. Active Backup for Microsoft 365
It is configured with 4 x 16 TB Seagate ST16000NT001-3LV101 hard drives in an SHR2 Storage Pool. I have created 2 x ~14 TB Volumes, as in the past, I have had issues with Windows accessing spaces > 16 TB with normal size file blocks. I also have 2 x 1 TB M.2 Samsung SSD 990 Pro drives to use for caching. I have followed your video and docs for using Synology HDD db so that I can use the Samsung drives for caching. I have set the script to run automatically on reboots.
My questions are as follows:
- Can you use SSD Caching across two separate volumes?
- If I were to use a read/write cache with pinned btrfs metadata (for Active Backup), how risky is it if the ssd array failed? Is all lost?
- all advice appreciated

[ed]

You can use SSD cache across multiple volumes as long as they live inside the same storage pool, which is true in your case since both of your fourteen terabyte volumes sit on the same SHR2 pool. Synology binds cache to the pool rather than individual volumes so both volumes benefit from the same cache layer without any extra configuration. With small file workloads and Active Backup running block level operations you will definitely see gains.

About using read write cache with pinned BTRFS metadata, it is faster but you need to be aware of risk. With write caching, any data waiting to be written to the HDD array sits temporarily on the SSDs. If one of your NVMe drives fails and you only run them as a single cache device, you can lose whatever was queued plus the metadata changes pinned to that SSD. If you run the SSDs as a redundant cache pair, the risk is far lower, because Synology mirrors the cache writes. Even then, the same rule applies as with any write back cache system: if the whole SSD cache array goes down suddenly, you may lose the unwritten data but you do not lose the entire pool. The underlying SHR2 array remains intact. In most business deployments the acceptable balance is a read write mirrored cache plus metadata pinning, especially when Active Backup runs constantly and touches many small blocks.

If you want to avoid any risk entirely you can run read only cache which never holds uncommitted data. But with your two NVMe 990 Pros and SHR2 on large HDDs the mirrored read write cache will give you the smoothest experience for your twenty user environment.

P.S. Regarding compatibility - you may need to use HDD script.