System and method of dynamically adjusting virtual machines for a workload

Patent metadata

Field	Value
Publication number	US-12423160-B2
Application number	US-202217806191-A
Country	US
Kind code	B2
Filing date	Jun 9, 2022
Priority date	Jun 9, 2022
Publication date	Sep 23, 2025
Grant date	Sep 23, 2025

Abstract

A method for dynamically adjusting a number of virtual machines for a workload, includes: receiving a probability indicator for each of a plurality of N sequential stages, where N is a natural number greater than 1, of a likelihood that a virtual machine assigned to a workload will be evicted during the N sequential stages; predicting a target number of virtual machines to configure in a current stage for a subsequent stage from among the plurality of N sequential stages based on the probability indicator, a target capacity for the workload, and a current price for maintaining a virtual machine; and configuring a number of virtual machines for the workload during the current stage based on the target number to be loaded for the workload for the subsequent stage.

First claim

What is claimed is:

1. A method for dynamically adjusting a number of virtual machines for a workload, comprising:
receiving a set of probability indicators for a plurality of N sequential stages of a likelihood that a virtual machine assigned to a workload will be evicted during the N sequential stages, wherein the plurality of N sequential stages comprises a current stage, an (N−1)-th stage after the current stage as the subsequent stage, and an N-th stage after the (N−1)-th stage as a last stage from among the plurality of N sequential stages, and wherein the set of probability indicators comprises:
a first probability indicator of a likelihood that a virtual machine assigned to the workload will be evicted during the current stage;
an (N−1)-th probability indicator of a likelihood that a virtual machine assigned to the workload will be evicted during the subsequent stage; and
an N-th probability indicator of a likelihood that a virtual machine assigned to the workload will be evicted during the last stage;

predicting a target number of virtual machines to configure in the current stage for the subsequent stage from among the plurality of N sequential stages based on the set of probability indicators, a target capacity for the workload, and a current price for maintaining a virtual machine; and
configuring a number of virtual machines for the workload during the current stage based on the target number to be loaded for the workload for the subsequent stage.

2. The method of claim 1 , wherein the N sequential stages correspond to different time horizons from each other.

3. The method of claim 2 , wherein a length of each of the N sequential stages corresponds to a configuration time of a virtual machine, and the configuring of the number of virtual machines for the workload takes a length of one stage to complete.

4. The method of claim 1 , wherein the predicting of the target number comprises:
constraining, based on a target capacity for the workload, the target number based on an anticipated changed capacity estimated for each of the plurality of N sequential stages;
constraining, based on a budget allocated for the workload, the target number based on a cost associated with the anticipated changed capacity of each of the current stage and the subsequent stage; and
calculating, based on a cost function, a final value for the target number that minimizes the cost function, while satisfying at least the constraints based on the target capacity and the budget.

5. The method of claim 4 , wherein the final value for the target number is determined according to an integer linear program, and the constraints based on the target capacity and the budget are inputs constraining the target number in the cost function.

6. The method of claim 4 , wherein to estimate the anticipated changed capacity for each of the plurality of N sequential stages, the method comprises:
determining a current number of virtual machines assigned to the workload at a start of the current stage;
estimating the anticipated changed capacity for the current stage that satisfy the target capacity, the anticipated changed capacity for the current stage corresponding to a change in the current number of virtual machines at an end of the current stage based on the first probability indicator and a value of the target number that is constrained by the target capacity;
estimating an anticipated changed capacity for the subsequent stage that satisfy the target capacity based on the anticipated changed capacity for the current stage, the (N−1)-th probability indicator, and a value of a potential number of virtual machines to configure during the subsequent stage that is constrained by the target capacity; and
estimating an anticipated changed capacity for the last stage that satisfy the target capacity based on the anticipated changed capacity of the current stage, the anticipated changed capacity of the subsequent stage, and the N-th probability indicator,
wherein the anticipated changed capacity for the last stage further constrains the values of the target number and the potential number based on the target capacity.

7. The method of claim 6 , wherein to determine the cost associated with the anticipated changed capacity of each of the current stage and the subsequent stage, the method comprises:
determining a current price for maintaining a virtual machine for the workload;
calculating a current cost associated with the anticipated changed capacity of the current stage based on the current price, and constrained by the budget; and
calculating a current cost associated with the anticipated changed capacity of the subsequent stage based on the current price, and constrained by the budget.

8. The method of claim 7 , wherein the target number of virtual machines comprises a first virtual machine, and a second virtual machine having a higher cost than that of the first virtual machine, and
wherein the budget corresponds to the target capacity multiplied by the higher cost of the second virtual machine.

9. The method of claim 4 , wherein the final value for the target number is further constrained according to a threshold limit of a maximum number of virtual machines that may be added in one stage, and a threshold limit of a total number of virtual machines that may be assigned to the workload at any given time.

10. A system for dynamically adjusting a number of virtual machines assigned to a workload, comprising:
a processor; and
memory comprising instructions that, when executed by the processor, cause the processor to:
receive a set of probability indicators for a plurality of N sequential stages of a likelihood that a virtual machine assigned to a workload will be evicted during the N sequential stages, wherein the plurality of N sequential stages comprises a current stage, an (N−1)-th stage after the current stage as the subsequent stage, and an N-th stage after the (N−1)-th stage as a last stage from among the plurality of N sequential stages, and wherein the set of probability indicators comprises:
a first probability indicator of a likelihood that a virtual machine assigned to the workload will be evicted during the current stage;
an (N−1)-th probability indicator of a likelihood that a virtual machine assigned to the workload will be evicted during the subsequent stage; and
an N-th probability indicator of a likelihood that a virtual machine assigned to the workload will be evicted during the last stage;

predict a target number of virtual machines to configure in the current stage for the subsequent stage from among the plurality of N sequential stages based on set of the probability indicators, a target capacity for the workload, and a current price for maintaining a virtual machine; and
configure a number of virtual machines for the workload during the current stage based on the target number to be loaded for the workload for the subsequent stage.

11. The system of claim 10 , wherein the N sequential stages correspond to different time horizons from each other.

12. The system of claim 11 , wherein a length of each of the N sequential stages corresponds to a configuration time of a virtual machine, and the configuring of the number of virtual machines for the workload takes a length of one stage to complete.

13. The system of claim 10 , wherein to predict the target number, the instructions further cause the processor to:
constrain, based on a target capacity for the workload, the target number based on an anticipated changed capacity estimated for each of the plurality of N sequential stages;
constrain, based on a budget allocated for the workload, the targe

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Ram Soumya
• Stephenson Preston Tapley
• Fischer Alexander David
• Sayed Mahmoud
• Minneker Robert Edward
• Vilarinho Eli Cortez Custodio
• Vieira Frujeri Felipe
• Goiri Presa Inigo
• Panjwani Sidhanth M
• Wang Yandan
• Couturier Camille Jean
• Zhang Jue
• Yang Fangkai
• Qin Si
• Lin Qingwei
• Bansal Chetan
• Pang Bowen
• Gupta Vivek

Classifications

• G06F2209/503

  Resource availability · CPC title

• G06F2209/5022

  Workload threshold · CPC title

• G06F9/5083

  Techniques for rebalancing the load in a distributed system · CPC title

• G06F9/505 Primary

  considering the load · CPC title

• G06F9/5033

  considering data affinity · CPC title