NVMe

In May 2022 it's being quite well behaved. They're happy everything is good and It's a much more efficient database. We're going from running 177 Cassandra knows down to 72 celadibino of course these Celadibino's are larger in size compared to the 77 177 Cassandra nodes right And a seller has nine terabyte compared to Cassandra which has four terabyte nodes Right Our tail latency improved from 40 to 125 millisecond down to 15 millisecond from 125 to 15 that is so chill that is so cool, Insert performance went from 5 millisecond to 70 down to a steady five This tells me that they did something with the disks The I they speed up they sped up their IO