Hi.
I read through your guides implementing 2 node cluster on top of Storage Spaces instead of hardware raid. Obviosly, most exciting feature in such design is tiering. I will try to validate such design in lab, but i need to ask a few questions first (before i start purchasing hardware).
By now, it's seems that 4 TB total storage capacity will be sufficient for our needs. I guess that there are much less than 1TB of hot data. So:
1TB SSD tier - 1 column, mirror resiliency - 2 x 1 TB SSD
6 TB HDD tier - 2 column, mirror resiliency - 4 x 3 TB HDD (i already have this disks)
2 port 10G Ethernet adapters on nodes (1 port Starwind sync, 1 port iSCSI)
First question! Hot/cold tier capacity ratio is significant less than reccomended by MS (30%). But if hot data less than 1TB, i think it is not real problem.
I plan to place two ha-device on single storage space disk.
Second question! 10G Ethernet for sync traffic between nodes will be enough in such desing? 1 SSD = ~500 MB/s read/write, 4xSSD = 2GB/s, and this is greater than 10G Ethernet throughput. I believe, such calculations are slightly naive. Will be a 10G network become a bottleneck?
Third and last question. What about flash memory wearing? Is there some addition load by starwind sync or Storage Spaces tiering? Can i just assume, that 1GB of data written to endpoint vhdx will be written on each 4 SSDs?
The Latest Gartner® Magic Quadrant™Hyperconverged Infrastructure Software