If SSD and SAS drives daisy chained together, and shared the same stack in a shelve connection, why is it not recommeneded?
I heard it should be separated and using its own port, but what issue it could cause and why?
Thanks for your sharing.
I work for NetApp and have done many storage cluster architectures, and hope to answer this for you!
It is not a matter of "not recommended" but slight change - It is recommended if you mix the shelf types that you keep the SSDs at the beginning of the loop or at the end of the loop.
This is because we purchase SSD shelves for the best performance, so you want the controller to have the most direct-direct-direct access to the SSD shelf. Not Loop1 being SAS -> SSD -> SAS.
To make the message more simple, we say if you have extra SAS ports then to just dedicate the loop to SSD, since SSD IOPS can really push a SAS HBA hard.
Hope this helps clarify!
There are also tighter limits on stack depth for stacks containiing SSD. You don't want to go over 4 shelves in the stack if it contains SSD. Should be covered in the storage subsystem faq.