We are planning to get
HPE DL380 Gen9 8SFF CTO Servers with
HP Smart Array P440ar/2G FIO Controller to setup a MongoDB cluster which will store huge amounts of data. (Assume growing non-stop)
Cluster will have 3 servers initially and will add more servers when needed for sharding.
Each server will have 16 * 10K SAS drives:
HPE 1.8TB SAS 10K SFF SC 512e HDD
MongoDB documentation suggests to have storage backed by RAID-10.
It does feel like waste of raw space considering RAID-10 yields only %50 usable size.
RAID-50 will give remarkably more storage, but only one drive can fail per sub-array, in case of re-build, performance degradation can be bad.
RAID-60 will use an additional disk for parity per sub-array, will have less performance, less space but more resilient.
My options with 16 * 1.8TB disks, using this calculator:
RAID-10: 13.1 TB
RAID-50: 22.9 TB
RAID-60: 19.6 TB
I've never dealt with an array of this size, considering challenges for a volume this large:
- RAID-60 costs additional disks but is performance penalty considerably less during degradation/rebuild?
- Any other pros/cons I should consider?