I have a couple of questions regarding mixing disk types into one aggregate. We will soon get a new FAS6280 clustered config with 3 kinds of disks. • 36 x 100GB SSD DISKS • 72 x 600 GB SAS 10000 RPM • 72 x 1 TB SATA 72000 RPM
This is the first time that I will get a Netapp config with different kinds of disks under the same controllers. I'm wondering what would be the best way to create the aggregate(s).
Would it be wise to combine the different disk types into one aggregate, or should we create one aggregate per disk type ?
I guess it should be possible to use all disks in one aggregate as long as we create RAID groups per disk type and do not mix disks types in the RAID groups right?
If it is possible, would there be a benefit in mixing the different disk types into one aggregate? We would like to have the possibility to migrate a volume between Tiers later as we see fit, for example moving a volume from SAS to SATA, from SAS to SSD, etc. without any downtime for the hosts that are using the volume.
If we decide to add all disk types to the same aggregate, would it be possible to do this?
Would there be any disadvantages to doing this ? Or am I completely wrong here and should we create 3 separate aggregates instead ?
If you put 10k SAS and 7,2k SATA to same aggregate, then SAS disks operate as slow as SATA disks, so you would just lose in this case. If it clustered ontap, you can move volume from SATA aggregate to SAS aggregate NDU.
Data ONTAP won't allow a mix of SAS and SATA in the same aggregate so the good news is it prevents the ability to have degraded drives like this... FlashPools (formerly hybrid aggregates) do allow mixing SSD with HDD (either SAS or SATA though or you can create a separate aggregate with all SSD....definitely a good point above on what the configuration was sized for and the layout in concept for the workloads.
We will probably do it the following way: One controller with a SATA aggregate + SSD's as cache One controller with a SAS aggregate. Maybe we will add some SSD's to the SAS aggregate as cache as well, or maybe we will use all SSDs on that controller for a separate SSD tier for high-performance databases. Still thinking about that one.
For the transparent moving of volumes between aggregates, I guess we can use the “Cluster-Mode Transparent Volume Move” feature.
I have one more question about adding the SSD's to the aggregate as cache: When we would add some SSD disks to the SAS aggregate for example, do they really become a part of the array? or will they be a special kind of disk that can be removed from the array again when needed without too much hassle ? We would like to play around a bit with this, add some SSD's to the SAS aggregate and then evaluate the performance gain, and maybe remove them again later in case we decide to use them for a separate SSD aggregate instead. Is this possible, or do the SSD's have to stay in the SAS aggregate forever after adding them as "cache" disks ?