it _is_ cheaper to run. But thats only part of the lifecycle. Where is the tooling to make the HA DB cluster? where is the tooling that securely and reliably d

it _is_ cheaper to run. But thats only part of the lifecycle. Where is the tooling to make the HA DB cluster? where is the tooling that securely and reliably does point in time recovery with a minute resolution for the last _n_ months? Where is the tooling to measure the health of the underlying FS, OS and hardware. What about scaling, what happens when you change the schema and shit gets slow, how long does it take to re-tool the scripts to make it work on new hardware? All of these will have to be built by your company, or bought in. Who tests them? how do you get support when shit goes wrong? For sure, it can be cheaper for the whole lifecycle, but often its not. Look I love real steel, I looked after a large cluster, but the prospect of making changes to the three-node tungsten, or creating a new one, gave me the fear.