Adapting a pre-trained distributed resource predictive model to a target distributed computing environment

• US 10,691,491 B2
• Filed: 10/19/2016
• Issued: 06/23/2020
• Est. Priority Date: 10/19/2016
• Status: Active Grant

• Current Assignee: Nutanix Incorporated
• Original Assignee: Nutanix Incorporated
• Inventors: Nagpal, Abhinay, Ramesh, Aditya, Shukla, Himanshu, Singh, Rahul
• Primary Examiner(s): Sun, Charlie
• Application Number:
  US15/298,149
• Publication Number:
  US 20200034197A1
• Time in Patent Office: 1,343 Days
• CPC Class Codes: G06F 9/50 Allocation of resources, e.g. of the central processing unit [CPU]
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Systems for distributed resource system management. A first computing system operates in a first computing environment. A predictive model is trained in the first computing environment to form a trained resource performance predictive model that comprises a set of trained model parameters to capture at least computing and storage IO parameters that are responsive to execution of one or more workloads that consume computing and storage resources in the first computing environment. When the trained resource performance predictive model is deployed to a second computing environment, various computing system configuration differences, and/or workload differences and/or other differences between the first computing environment and the second computing environment are detected and measured. Responsive to the detected differences and/or measurements, some of the trained resource performance predictive model parameters are modified to adapt the trained resource performance predictive model to any of the detected and/or measured characteristics of the second computing environment.

Claims

1. A method, comprising:

training a model into a trained model, wherein the model comprises a set of trained model parameters to capture resource consumption of computing and storage resources of a workload in a first computing environment, and the first computing environment is characterized by a first configuration;

detecting a difference between the first computing environment and a second computing environment characterized by a second configuration, wherein

the first configuration comprises at least a portion of the second configuration so that the first computing environment in which the model is trained comprises a computing node that is also in the second computing environment into which the trained model is to be deployed; and

deploying the trained model to the second computing environment at least by adapting the trained model to the second computing environment, wherein adapting the trained model comprises modifying the set of model parameters based at least in part on the difference prior to conclusion of adapting the trained model to the second computing environment.

2. The method of claim 1, wherein the difference comprises a configuration difference between a target system configuration and a training system configuration.

3. The method of claim 1, wherein training the model comprises:

determining a storage IO performance characteristic at least by executing a training workload schedule in the first computing environment, wherein the trained model is adapted to the second computing environment based further at least in part upon the storage IO performance characteristic.

4. The method of claim 1, wherein the first computing environment corresponds to a training workload schedule, the second computing environment corresponds to a target workload schedule, and the training workload schedule comprises at least a portion of the target workload schedule.

5. The method of claim 1, wherein the first computing environment corresponds to a training workload schedule, the second computing environment corresponds to a target workload schedule, and the difference comprises a workload schedule difference between the target workload schedule and the training workload schedule.

6. The method of claim 1, wherein modifying the set of model parameters is performed prior to commencing a learning phase during which a relationship between an input and an output of the trained model is determined for the second computing environment in the second computing environment.

7. The method of claim 1, wherein the set of model parameters characterizes a correlation between a stimulus and a response in the first computing environment.

8. The method of claim 7, wherein modifying the set of model parameters comprises at least one of removing, attenuating, amplifying, or time-shifting a correlation that relates a stimulus to a response in the first computing environment.

9. The method of claim 1, wherein the a state of the trained model in the second computing environment is characterized by the set of model parameters, which has been modified, at the conclusion of adapting the trained model.

10. A non-transitory computer readable medium having stored thereon a sequence of instructions which, when stored in memory and executed by a processor, causes the processor to perform a set of acts, the set of acts comprising:

training a model into a trained model, wherein the model comprises a set of model parameters to capture resource consumption of computing and storage resources of a workload in a first computing environment, and the first computing environment is characterized by a first configuration;

detecting a difference between the first computing environment and a second computing environment characterized by a second configuration, wherein

the first configuration comprises at least a portion of the second configuration so that the first computing environment in which the model is trained comprises a computing node that is also in the second computing environment into which the trained model is to be deployed; and

deploying the trained model to the second computing environment, at least by adapting the trained model to the second computing environment, wherein adapting the trained model comprises modifying the set of model parameters based at least in part on the difference prior to conclusion of adapting the trained model to the second computing environment.

11. The non-transitory computer readable medium of claim 10, wherein the difference comprises a configuration difference between a target system configuration and a training system configuration.

12. The non-transitory computer readable medium of claim 10, wherein training the model comprises:

determining a storage IO performance characteristic at least by executing the training workload schedule in the first computing environment, wherein the trained model is adapted to the second computing environment based further at least in part upon the storage IO performance characteristic.

13. The non-transitory computer readable medium of claim 10, wherein the first computing environment corresponds to a training workload schedule, the second computing environment corresponds to a target workload schedule, and the training workload schedule comprises at least a portion of a target workload schedule.

14. The non-transitory computer readable medium of claim 10, wherein the first computing environment corresponds to a training workload schedule, the second computing environment corresponds to a target workload schedule, and the difference comprises a workload schedule difference between a target workload schedule and the training workload schedule.

15. The non-transitory computer readable medium of claim 10, wherein modifying the set of model parameters is performed prior to commencing a learning phase during which a relationship between an input and an output of the trained model is determined for the second computing environment.

16. The non-transitory computer readable medium of claim 10, wherein the set of model parameters characterizes a correlation between a stimulus and a response in the first computing environment.

17. A system, comprising:

a storage medium having stored thereon a sequence of instructions; and

a processor that executes the instructions to cause the processor to perform a set of acts, the set of acts comprising,

training a model into a trained model, wherein the model comprises a set of model parameters to capture resource consumption of computing and storage resources of a workload in a first computing environment, and the first computing environment is characterized by a first configuration;

detecting a difference between the first computing environment and a second computing environment characterized by a second configuration, wherein

the first configuration comprises at least a portion of the second configuration so that the first computing environment in which the model is trained comprises a computing node that is also in the second computing environment into which the trained model is to be deployed; and

deploying the trained model to the second computing environment, wherein at least by adapting the trained model to the second computing environment, wherein adapting the model comprises modifying the set of trained model parameters based at least in part on the difference prior to conclusion of adapting the trained model to the second computing environment.

18. The system of claim 17, wherein the difference comprises a configuration difference between a target system configuration and a training system configuration.

19. The system of claim 17, wherein training the model comprises:

determining a storage IO performance characteristic at least by executing a training workload schedule in the first computing environment, wherein the trained model is adapted to the second computing environment based at least in part upon the storage IO performance characteristic.

20. The non-transitory computer readable medium of claim 17, wherein modifying the set of model parameters is performed prior to commencing a learning phase during which a relationship between an input and an output of the trained model is determined for the second computing environment.