| 摘要 |
A method for mirroring virtual machines (202) from a primary host (100) to a secondary host (102) uses checkpoints and includes tracking changes 404 for each of a plurality of memory pages (304) and processor states for one or more primary host virtual machines. Responsive to an occurrence of a checkpoint 408, the primary host virtual machines are stopped. A determination 412 is made if each of the memory pages is frequently changed. In response to the memory page being frequently changed, the frequently changed. memory page is marked as being writeable 422 and copied 423 to a buffer. In response to the memory page being infrequently changed, the infrequently changed memory page is marked as being read only 414. The one or more primary host virtual machines are resumed 415. A copy of the memory pages, the buffer, and changes to the processor states are transmitted to the secondary host. Tracking frequently used memory pages reduces the occurrence of memory page protection faults and the overhead caused by interrupting the virtual machine to service memory page faults for frequently-accessed pages is reduced or eliminated.
|
