Is there a performance sizing considerations for IOPS/drive in a VxRAIL scenario, or do we follow VSAN recommendations? e.g. for a 10k SAS (typical 150iops RND), how does this work our when added in VxRAIL?
But, I don't think sizing using simply IOPS/Drive is a good ( or comprehensive ) strategy.
It will depends on how cache friendly the workload is, and the working set size, and both of those variables are very hard to determine.
On hybrid configurations the SSD Cache is divided in a read cache and a write buffer. If the working set fits the cache layer, the disks would not be on the data path, meaning the application will be acknowledged before the data is written to the disk ( capacity layer ).
It uses Mirroring or Erasure Coding, and protects the objects ( i.e VMDKs ) instead of disks.
Hybrid configurations can choose from FTT=0 ( no protection ) to FTT=3 ( Triple Mirroring ).
All Flash configurations also have the Erasure Coding option, which offers a protection level similar to RAID5 or RAID6. It is managed by the policy "Fault Tolerance Method".
GustavoEMC
15 Posts
3164
2
Posted August 18th, 2016 09:00
Hi,
On hybrid configurations the usable area is roughly 48/49% of the raw capacity when using FTT=1.
So, in a 4 node with 5 X 1.2TB disks per node, you have ~11TB usable. If using 4 X 1.2TB disks per node you have ~9TB usable.
Does this answer your question ?
Regards,