That was a very beautiful explanation that creates a good intuition of benefits and limits of storage disagregated systems. Writing here a very very simplified explanation for my future reference;
Naive remote storage 15x-20x slower, Log as db P2 design 20% overhead, Multiple Version Storage (P3) for access from multiple nodes avoiding "torn page" using different version of the pages 40% overhead
That was a very beautiful explanation that creates a good intuition of benefits and limits of storage disagregated systems. Writing here a very very simplified explanation for my future reference;
Naive remote storage 15x-20x slower, Log as db P2 design 20% overhead, Multiple Version Storage (P3) for access from multiple nodes avoiding "torn page" using different version of the pages 40% overhead