Oh, interesting on the memory! Large, complex, DBs such as TPC-DS SF30 and higher thrive on memory, especially when running concurrent users. That is where JUMBO comes into its own vesus STANDARD.
Once I've finished the comparison with 2024, I'll go back to TPC-DS as it provides a superb platform for testing, especially when the SF is > 30 or so. I managed up to SF250 once before, but will return to that when I've finshed the current work.
P.S. some very early results after installing 1.2.2 and latest MD show that instances seem to work even better than before, especially when specifying separte hings per replica and adding the dbinstance_inactivity_ttl=10 to ensure the connections remain with the same server/duckling.
.png){kind=logo}
