2015-09-24 07:02 AM - edited 2015-09-24 07:36 AM
We just purchased a new FAS2520 with a 12 disk shelf. We have 8x4TB drives, and 4x186GB SSDs. We are using OnTap 8.3P2 and will be upgrading to 8.3.1 during the new ADP configuration. We were unaware of ADP. So our initial setup was a simple 50/50 split were we ran into issues obviously when trying to create aggregates a min of 5 discs is needed. We obviously have 4 SAS and 2 SSD per node.
We are thinking of going towards an ADP route and most likely Active / Passive where we assign all discs to 1 node to maximize usable space. We would like to have about 15-16TB of backups and whatever is leftover for non-critical internal VMs 1-3TB is perfectly fine.
Can anyone recommend an ADP setup? Reading up we are seeing that 5 of our 8 SAS drives would have to be DATA while 2 will be parity, and 1 for spare? How many disc failures will this allow for 1, 2? I am assuming we need 1 parity disc per node and can't do a 6+1+1 setup it will need to be 5+2+1? If we went with a 5+2+1 setup how much usable space would we have to work with? And finally what should we do with our SSD's? Create 1 large flash pool between the 4?
2015-09-25 10:31 AM
With 8 SATA and 4 SSD, you're likely looking at a single FlashPool aggregate, and I would assume it was sold to you as such.
ADP means not having to dedicate drives for mroot, but beyond that, you can think of this as an aggregate creation discussion.
The main decision you need to make is to choose between capacity/performance and spares on your 8 SATA drives.
Possible aggregate configurations would be:
8 (rg0 = 6 data, 2 parity), 0 spares = 24TB RAW
7 (rg0 = 5 data, 2 parity), 1 spares = 20TB RAW
6 (rg0 = 4 data, 2 parity), 2 spares = 16TB RAW
5 (rg0 = 3 data, 2 parity), 3 spares = 12TB RAW
5 data, 2 parity, 1 spare is probably the most common option; assuming a 90% ceiling, 20TB RAW is about 14TB usable.
Fault tolerance, being RAID-DP, allows for two, simultaneous drive failures per raidgroup without data loss.
Parity drive requirements are per raidgroup, not per node, so all your drives would be owned by one node.
Once your SATA drive aggreate is created, you would then add your SSD's to make it a FlashPool.