Work out vm memory ballooning instantly with clear inputs, formula shown and shareable results.
The balloon driver inflates inside the guest so the guest's own kernel chooses which pages to release, which is far better than the hypervisor swapping blindly. It works only while guests hold genuinely idle memory: once active demand exceeds host RAM, ballooning forces guest swapping and performance collapses.
Ballooning requirement
available for guests = host memory - hypervisor reserve; reclaim = allocated - available; active demand = allocated x active share
The guest kernel knows which pages are cold. The hypervisor does not, so host swapping frequently pages out memory the guest is about to use.
Reserving memory rather than disabling the driver is the better approach, so the hypervisor never targets that VM for reclamation.