Hello guys,
Tried to find the answer in forum, found several interesting cases and discussions about L2 cache and MS Storage Spaces,
but somehow nothing that would answer my question. Comments or recommendations needed.
Suggested environment (although it's not so important as question is more architectural):
Two identical nodes in hyper converged cluster running Windows Server 2016 Datacenter:
hardware RAID controller available
2 SSDs and 4 HDDs dedicated for storage (with capacity ratio SSD/HDD – 1:10)
one NIC with 2x10Gb and MB integrated - 4x1Gb
------------------------------------------------------------------
1x10Gb – StarWind sync, directly connected;
1x10Gb – StarWind iSCSI targets, directly connected;
3x1Gb – Cluster live migration, directly connected;
1x1Gb – Management, plugged into switch.
-------------------------------------------------------------------
As I understand it’s advised to use hardware RAID10, if available, for the storage.
The question is which way would be more beneficial to use SSDs to improve HDDs’ performance:
1) Configure SSD RAID10 disk and HDD RAID10 disk using RAID controller and use SSD RAID10 disk as L2 cache for StarWind flat image disks;
2) Configure SSD RAID10 disk and HDD RAID10 disk using RAID controller, pool both disks to storage spaces, create simple tiered virtual disk
for StarWind flat images and let windows distribute the data between tiers depending on usage frequency and refuse using L2 cache.
First choice, I suggest, will improve reads, while second – reads and writes, but adding additional software layer will mean additional processing loads.
Also i didn't investigate on which level MS Storage Spaces work distributing data between tiers, so possibly, if it's not a block level and OS will see only
a single StarWind image file, it will be unable to use tiers?
I'm quite fresh as concerns storage matters, so sorry in advance if asking obvious or dumb things.
The Latest Gartner® Magic Quadrant™Hyperconverged Infrastructure Software