Did you consider using the same network storage for the bare metal tests to measure the relative overhead introduced like the title "Base Metal vs Virtual" suggests?
Is it possible the jitter measured in ops/sec was caused by congestion on the network storage tier?
Did you consider using the same network storage for the bare metal tests to measure the relative overhead introduced like the title "Base Metal vs Virtual" suggests?
Is it possible the jitter measured in ops/sec was caused by congestion on the network storage tier?