Dynamically selecting optimal instance type for disaster recovery in the cloud
The selection of an optimal restore instance type based on a customer's speed/cost tradeoff resolution is disclosed. Aspects of the anticipated completion time to complete the recovery and completion cost to perform the recovery may be extrapolated based on a baseline or test recovery and/or actual recovery times and costs. An automated restore activity may be performed on a baseline test VM of a predefined size using different restore instance types. An optimal restore instance type is used to form worker VMs that perform the recovery operations.
1. A method, comprising:
restoring a test virtual machine (VM) at a target site using multiple worker VMs that are based on a multiple VM restore instance types, wherein, for each VM restore instance type, results of restoring the test VM is associated with a completion time and a completion price;
extrapolating, for each restore instance type, the completion time and the completion price based on the results of restoring the test VM and actual results of test restores or actual restores performed by other entities;
generating, for each of the VM restore instance types, a baseline vector based on the restoration of the test VM, the baseline vector identifying the VM restore instance type;
generating a second vector based on the baseline vector for each of the baseline vectors, wherein each of the second vectors includes an extrapolated completion time and an extrapolated completion price; and
ranking the second vectors.
2. The method as recited in claim 1 , wherein the baseline vector is generic, with respect to a VM restore instance type, to a plurality of VMs at the production site.
3. The method as recited in claim 1 , wherein each of the second vectors is specific to a particular VM restore instance type.
4. The method as recited in claim 1 , wherein extrapolating is based on data including one or more of VM size, disk type, restore instance type, the completion time and the completion price.
5. The method as recited in claim 4 , wherein information in the second vector concerning the time and cost to restore the VM identified in that second vector is derived, respectively, from the time needed to restore the test VM at the cloud storage site, and the cost to restore the test VM at the cloud storage site.
6. The method as recited in claim 1 , wherein the second vectors are ranked based on a customer speed/cost tradeoff resolution policy.
7. The method as recited in claim 1 , further comprising restoring, at the target site, the production site VM identified in the second vector using a worker VM at that is based on a restore instance type identified in the second vector.
8. The method as recited in claim 1 , further comprising performing a search on less than all of the restore instance types to identify an optimal restore instance type and performing a restore operation with a restore VM based on the optimal restore instance type.
9. The method of claim 8 , further comprising perform the search for each tier of bandwidths associated with the restore instance types.
10. The method of claim 1 , wherein extrapolating, for each restore instance type, includes performing one or more of a linear extrapolation, a non-linear extrapolation, or a least squares linear regression.
11. A non-transitory storage medium having stored therein computer-executable instructions which, when executed by one or more hardware processors, perform the operations:
restoring a test virtual machine (VM) at a target site using a multiple worker VMs that are based on multiple VM restore instance types, wherein, for each VM restore instance type, results of restoring the test VM is associated with a completion time and a completion price;
extrapolating, for each restore instance type, the completion time and the completion price based on the results of restoring the test VM and actual results performed by other entities;
generating, for each of the VM restore instance types, a baseline vector based on the restoration of the test VM, the baseline vector identifying the VM restore instance type;
generating a second vector based on the baseline vector for each of the baseline vectors, wherein each of the second vectors includes an extrapolated completion time and an extrapolated completion price; and
ranking the second vectors.
12. The non-transitory storage medium as recited in claim 11 , wherein the baseline vector is generic, with respect to a VM restore instance type, to a plurality of VMs at the production site.
13. The non-transitory storage medium as recited in claim 11 , wherein each of the second vectors is specific to a particular VM restore instance type.
14. The non-transitory storage medium as recited in claim 11 , wherein extrapolating is based on data including one or more of VM size, disk type, restore instance type, the completion time and the completion price.
15. The non-transitory storage medium as recited in claim 11 , wherein information in the second vector concerning the time and cost to restore the VM identified in that second vector is derived, respectively, from the time needed to restore the test VM at the cloud storage site, and the cost to restore the test VM at the cloud storage site.
16. The non-transitory storage medium as recited in claim 11 , wherein the second vectors are ranked based on a customer speed/cost tradeoff resolution policy.
17. The non-transitory storage medium as recited in claim 11 , further comprising restoring, at the target site, the production site VM identified in the second vector using a worker VM at that is based on a restore instance type identified in the second vector.
18. The non-transitory storage medium as recited in claim 11 , further comprising performing a search on less than all of the restore instance types to identify an optimal restore instance type and performing a restore operation with a restore VM based on the optimal restore instance type.
19. The non-transitory storage medium as recited in claim 11 , further comprising perform the search for each tier of bandwidths associated with the restore instance types.
20. The non-transitory storage medium as recited in claim 11 , wherein extrapolating, for each restore instance type, includes performing one or more of a linear extrapolation, a non-linear extrapolation, or a least squares linear regression.