Memory ballooning is a memory management technique that allows a physical host to take advantage of unused memory on its guest virtual machines (VMs).
When a host OS is short on physical memory, the hypervisor polls balloon drivers installed on the guests to request available memory. The guests respond by using page-reclaiming algorithms to determine which pages are available and can be assigned to the balloon drivers. The process of assigning available pages to the driver is known as inflating the balloon. Releasing available pages is known as deflating the balloon. Continue reading