Understanding how SSD Cache works?

[Daisuke]

Hi guys,

Freenas documentation says that SSD cache devices only help if your dataset is larger than system RAM, but small enough that a significant percentage of it will fit on the SSD. In other words, if you have a 64GB disk and your dataset has 30GB used out of 5TB, it will help a lot. Please correct me if anyone thinks I got it wrong.

Personally, I have 1 dataset (mnt/nas) with 2 sub-datasets (/mnt/nas/unix and /mnt/nas/media). I store development data files like MySQL databases where the cache speed is beneficiary, into /unix dataset. That dataset is now 50GB, so it does fit into SSD without issues. I do not have a defined quota or reserved space for that dataset.

 

[Gnome]

I would assume you don't store the ACTUAL current database on your NAS?

Because ZFS is copy-on-write it isn't well suited to the database environment (constant small modifications). Ideally ZFS should store large files which are read often but modified seldom.

With that out of the way, the cache will store data that is being "hit" (read) the most.

If you dataset is so small you might even want to consider a SSD RAIDz? I think RAIDz of say 3x240GB OCZ Agility would before brilliantly (could probably come close to saturating 10GBe for sequential reads).