If they were making decisions based on RAM, they were almost certainly encountering real-world issues which prompted looking at their usage statistics in the first place, rather than just looking up their RAM usage metrics for funsies.
If they were making decisions based on RAM, they were almost certainly encountering real-world issues which prompted looking at their usage statistics in the first place, rather than just looking up their RAM usage metrics for funsies.
But to that point, having to explain to people why this 32GB machine with two processes using 10GB each can not, in fact, fit another 10GB process because the response times of the two existing processes relies heavily on that “unused” memory that is actually part of the working set of the machine. And also occasionally one of those services spikes to 14GB so what happens when that happens?
So in that regard showing that memory as used has advantages for avoiding arguments with people who are not equipped to have their half of the argument in a useful and meaningful way.
It makes sense to watch memory use but one should make sure to discount the amount used for disk caching from the total.