Back to results
US 12,585,502 B2Granted
Virtual machine management in data centers
AssigneeNVIDIA Corporation
InventorsAndrew Currid, Anshul Fadnavis, Chenghuan Jia, Ankit Agrawal
Life of this patent
Estimating the term…
Abstract
Apparatuses, systems, and techniques to determine metrics for paths connecting hardware components, select a plurality of groups of the hardware components based at least in part on the metrics, and perform at least a portion of a workload using a selected group of the plurality of groups.
Claims24
A method comprising:
- determining expected path performances for one or more paths between hardware components of a computing system connected by one or more connections;
- selecting a first group of at least two of the hardware components to perform a workload based at least in part on the expected path performances;
- executing a first virtual machine using the first group;
- mapping the at least two hardware components of the first group to the first virtual machine in a first order that represents relationships between the at least two hardware components of the first group;
- performing at least a portion of the workload using the first virtual machine;
- suspending performance of the workload before the workload is finished;
- preserving state information for the workload;
- selecting a different second group of at least two of the hardware components based at least in part on the different second group being compatible with the first group;
- executing a second virtual machine using the different second group;
- mapping the at least two hardware components of the different second group to the second virtual machine in a second order that represents relationships between the at least two hardware components of the different second group, wherein the second order is compatible with the first order;
- loading the state information into the second virtual machine;
- and
- resuming the performance of the workload using the second virtual machine.
10 more elements
A system comprising:
- one or more computing systems comprising hardware components connected by one or more connections;
- and one or more circuits to;
- identify a plurality of groups comprising at least two of the hardware components based at least on weights associated with paths between at least a portion of the hardware components along at least a portion of the one or more connections;
- select a first group of the plurality of groups to perform a workload;
- execute a first virtual machine using the first group;
- map the at least two hardware components of the first group to the first virtual machine in a first order that represents relationships between the at least two hardware components of the first group;
- perform at least a portion of the workload using the first virtual machine;
- suspend performance of the workload before the workload is finished;
- preserve state information for the workload;
- select a different second group of the plurality of groups based at least in part on the different second group being compatible with the first group;
- execute a second virtual machine using the different second group;
- map the at least two hardware components of the different second group to the second virtual machine in a second order that represents relationships between the at least two hardware components of the different second group, wherein the second order is compatible with the first order;
- load the state information into the second virtual machine;
- and
- resume the performance of the workload using the second virtual machine.
12 more elements
A processor comprising one or more circuits to:
- determine one or more metrics for paths connecting hardware components;
- select a plurality of groups of at least two of the hardware components based at least in part on the metrics;
- select a first group of the plurality of groups;
- execute a first virtual machine using the first group;
- map the at least two hardware components of the first group to the first virtual machine in a first order that represents relationships between the at least two hardware components of the first group;
- perform at least a portion of a workload using the first virtual machine;
- suspend performance of the workload before the workload is finished;
- preserve state information for the workload;
- select a different second group of the plurality of groups based at least in part on the different second group being compatible with the first group;
- execute a second virtual machine using the different second group;
- map the at least two hardware components of the different second group to the second virtual machine in a second order that represents relationships between the at least two hardware components of the different second group, wherein the second order is compatible with the first order;
- load the state information into the second virtual machine;
- and
- resume the performance of the workload using the second virtual machine.
11 more elements
Dependencies are inferred from the claim text.