In a multi-node configuration, does read cache and write buffer operate at the same IOPS as a RAM disk? I would think that read cache would operate at the speed of RAM disk because it does not need to sync to all nodes, but write buffer would not because it has to sync over the network to all nodes. Is this correct?
If so, then the network between the nodes would be a bottleneck for performance, correct? Does that mean FC would be a better choice than 10 Gb CAT6 ethernet, because latency is lower for FC? Is there something with lower latency than ethernet and FC?
I am trying to determine how to get the most performance out of SATA+RAM w/ StarWind.
The Latest Gartner® Magic Quadrant™Hyperconverged Infrastructure Software