Google patents flexible memory management for virtual machines

1. A method comprising: allocating, by a virtual machine monitor executing at a host computing system, an amount of memory for a guest virtual machine, the amount of memory being less than a maximum amount of memory the guest virtual machine is configured to utilize, wherein the host computing system provides an execution environment for the guest virtual machine; and after starting a boot process of the guest virtual machine: monitoring, by the virtual machine monitor, memory pressure of the guest virtual machine; responsive to an attempt to access a memory location of the guest virtual machine that is unallocated, receiving, by the virtual machine monitor, a memory access error that includes an indication of the memory location that was attempted to be accessed; responsive to determining, by the virtual machine monitor, that an amount of available memory of the host computing system satisfies a threshold amount of available memory: responsive to receiving the memory access error and based on the memory pressure, allocating, by the virtual machine monitor, additional memory for the guest virtual machine by reducing an amount of memory allocated to a virtual balloon, wherein the additional memory includes the memory location, and wherein the amount of memory allocated to the guest virtual machine is inversely proportional to the amount of memory allocated to the virtual balloon, wherein the additional memory is not allocated where the amount of available memory of the host computing system does not satisfy the threshold amount of available memory; and accessing, by the virtual machine monitor, the memory location.