distributed DBs provide seamless fault tolerance story usually. > Consider that a modern NVMe drive can quite easily deliver 1,000,000+ IOPS. There could be o

distributed DBs provide seamless fault tolerance story usually. > Consider that a modern NVMe drive can quite easily deliver 1,000,000+ IOPS. There could be other bottlenecks, for example I am consistently experiencing that linux kernel doesn't handle memory pages allocations fast enough once disk traffic hits few GB/s, because it does it in single thread.