Managing availability of virtual machines in cloud computing services

摘要

Recovery of virtual machines when one or more hosts fail includes identifying virtual machines running on the remaining functioning hosts. Some of the identified powered on virtual machines are suspended in favor of restarting some of the failed virtual machines from the failed host(s). A subsequent round of identifying virtual machines for suspension and virtual machines for restarting is performed. Virtual machines for suspension and restarting may be identified based on their associated “recovery time objective” (RTO) values or their “maximum number of RTO violations” value.

主权项

1. A method comprising:
instantiating a plurality of virtual machines on a plurality of virtual machine hosts (“hosts”), each virtual machine being associated with a recovery time objective (RTO) value and a maximum number of RTO violations (MRV) value, wherein identifying the M virtual machines is further based on an MRV value associated with each powered-on virtual machine; and responding to a failed host by restarting failed virtual machines from the failed host on remaining functioning hosts, including:
identifying M virtual machines from among a plurality of powered-on virtual machines running on the remaining functioning hosts based on an RTO value associated with each of the powered-on virtual machines;identifying N virtual machines from among the failed virtual machines based on an RTO value associated with each of the failed virtual machines;suspending the M virtual machines; andrestarting the N virtual machines on one or more of the remaining functioning hosts.