Intelligent scale-out federated restore
View Patent ↗A request is received to restore a set of virtual machines (VMs) to a cluster of nodes. The nodes are profiled to collect, for each node, a current load and health status. A participating node list is generated by excluding from the list any node having a current load that exceeds a threshold or having a health status of unavailable. For each participating node, a determination is made of a recommended number of VMs to assign to a node based on a current load of the node. A participating node and VM to restore is selected. The selected VM is assigned to the selected participating node. The selection and assignment is repeated until each participating node has been assigned its respective number of VMs to restore or all VMs to restore have been assigned. The participating nodes are instructed to begin restoring their assigned VMs.
1. A method comprising:
receiving a request to restore a plurality of virtual machines (VMs) to a cluster of nodes, each node comprising a plurality of resources for hosting one or more VMs;
profiling the cluster to collect, for each node, a current load on the plurality of resources and a health status;
generating a participating node list by excluding from the list any node having a current load that exceeds a threshold or having a health status of unavailable;
determining, for each node in the list, a recommended number of VMs to assign for restore based on a particular current load of a corresponding particular node;
selecting a node from the participating node list and a VM from the plurality of VMs to restore;
assigning the selected VM to the selected node to restore;
repeating the selecting and assigning until each node in the list has been assigned its respective recommended number of VMs to restore or all VMs from the plurality of VMs to restore have been assigned; and
instructing the nodes in the list to begin restoring their assigned VMs.
2. The method of claim 1 wherein the current load comprises memory usage, central processor unit (CPU) usage, and network usage.
3. The method of claim 1 wherein the determining, for each node in the list, a recommended number of VMs to assign for restore based on a particular current load of a corresponding particular node comprises:
calculating a remaining load capacity of the particular node as a function of memory usage, central processor unit (CPU) usage, and network usage of the particular node; and
calculating a particular recommended number of VMs to assign to the particular node as a function of the remaining load capacity of the particular node and a load demand of a VM,
wherein memory usage is assigned a higher weighted value than CPU usage and network usage, and CPU usage is assigned a higher weighted value than network usage.
4. The method of claim 1 comprising:
receiving from a first node in the participating node list an indication that the first node has completed restoration of VMs assigned to the first node; and
redistributing at least a subset of VMs initially assigned to a second node in the participating node list to restore to the first node to restore.
5. The method of claim 1 wherein a subset of nodes from the plurality of nodes to restore remain unassigned because each node in the participating node list has been assigned its respective recommended number of VMs to restore and the method comprises:
receiving from a first node in the participating node list an indication that the first node has completed restoration of VMs initially assigned to the first node; and
assigning to the first node at a least some of the subset of nodes to restore.
6. The method of claim 1 wherein nodes in the participating node list restore their assigned VMs in parallel with each other.
7. A system for performing federated restores, the system comprising:
a processor-based computer system configured to:
receive a request to restore a plurality of virtual machines (VMs) to a cluster of nodes, each node comprising a plurality of resources for hosting one or more VMs;
profile the cluster to collect, for each node, a current load on the plurality of resources and a health status;
generate a participating node list by excluding from the list any node having a current load that exceeds a threshold or having a health status of unavailable;
determine, for each node in the list, a recommended number of VMs to assign for restore based on a particular current load of a corresponding particular node;
select a node from the list and a VM from the plurality of VMs to restore;
assign the selected VM to the selected node to restore;
repeat the selection and assignment until each node in the list has been assigned its respective recommended number of VMs to restore or all VMs from the plurality of VMs to restore have been assigned; and
instruct the nodes in the list to begin restoring their assigned VMs.
8. The system of claim 7 wherein the current load comprises memory usage, central processor unit (CPU) usage, and network usage.
9. The system of claim 7 wherein the processor-based computer system is configured to:
calculate a remaining load capacity of the particular node as a function of memory usage, central processor unit (CPU) usage, and network usage of the particular node; and
calculate a particular recommended number of VMs to assign to the particular node as a function of the remaining load capacity of the particular node and a load demand of a VM,
wherein memory usage is assigned a higher weighted value than CPU usage and network usage, and CPU usage is assigned a higher weighted value than network usage.
10. The system of claim 7 wherein the processor-based computer system is configured to:
receive from a first node in the participating node list an indication that the first node has completed restoration of VMs assigned to the first node; and
redistribute at least a subset of VMs initially assigned to a second node in the participating node list to restore to the first node to restore.
11. The system of claim 7 wherein a subset of nodes from the plurality of nodes to restore remain unassigned because each node in the participating node list has been assigned its respective recommended number of VMs to restore and the processor-based computer system is configured to:
receive from a first node in the participating node list an indication that the first node has completed restoration of VMs initially assigned to the first node; and
assign to the first node at a least some of the subset of nodes to restore.
12. The system of claim 7 wherein nodes in the participating node list restore their assigned VMs in parallel with each other.
13. A computer program product, comprising a non-transitory computer-readable medium having a computer-readable program code embodied therein, the computer-readable program code adapted to be executed by one or more processors to implement a method comprising:
receiving a request to restore a plurality of virtual machines (VMs) from backup to a cluster of nodes, each node comprising a plurality of resources for hosting one or more VMs;
profiling the cluster to collect, for each node, a current load on the plurality of resources and a health status;
generating a participating node list by excluding from the list any node having a current load that exceeds a threshold or having a health status of unavailable;
determining, for each node in the list, a recommended number of VMs to assign for restore based on a particular current load of a corresponding particular node;
selecting a node from the participating node list and a VM from the plurality of VMs to restore;
assigning the selected VM to the selected node to restore;
repeating the selecting and assigning until each node in the list has been assigned its respective recommended number of VMs to restore or all VMs from the plurality of VMs to restore have been assigned; and
instructing the nodes in the list to begin restoring their assigned VMs.
14. The computer program product of claim 13 wherein the current load comprises memory usage, central processor unit (CPU) usage, and network usage.
15. The computer program product of claim 13 wherein the determining, for each node in the list, a recommended number of VMs to assign for restore based on a particular current load of a corresponding particular node comprises:
calculating a remaining load capacity of the particular node as a function of memory usage, central processor unit (CPU) usage, and network usage of the particular node; and
calculating a particular recommended number of VMs to assign to the particular node as a function of the remaining load capacity of the particular node and a load demand of a VM,
wherein memory usage is assigned a higher weighted value than CPU usage and network usage, and CPU usage is assigned a higher weighted value than network usage.
16. The computer program product of claim 13 wherein the method comprises:
receiving from a first node in the participating node list an indication that the first node has completed restoration of VMs assigned to the first node; and
redistributing at least a subset of VMs initially assigned to a second node in the participating node list to restore to the first node to restore.
17. The computer program product of claim 13 wherein a subset of nodes from the plurality of nodes to restore remain unassigned because each node in the participating node list has been assigned its respective recommended number of VMs to restore and the method comprises:
receiving from a first node in the participating node list an indication that the first node has completed restoration of VMs initially assigned to the first node; and
assigning to the first node at a least some of the subset of nodes to restore.
18. The computer program product of claim 13 wherein nodes in the participating node list restore their assigned VMs in parallel with each other.