We all know its essential to separate the workloads. small size block from large size block IO.
At least transmit them via different server ports. Otherwise small block IO will experience an elevated response time due to the fact that they will share the queue with large blocks and it takes more time to transmit a large block than a small one.
I wonder if this rule imposes the necessity to have separate HBA ASICs for different workloads as well?
Or ASIC works on the frame level and does not care what IO size is inside?
Frankly speaking I'm totally aware of correlation between the block size on disk level and frame size, so probably I miss a point here.
If anyone could make it clear ?