Initialization of resource allocation for a workload characterized using a regression model

Patent metadata

Field	Value
Publication number	US-11327801-B2
Application number	US-201916554897-A
Country	US
Kind code	B2
Filing date	Aug 29, 2019
Priority date	Aug 29, 2019
Publication date	May 10, 2022
Grant date	May 10, 2022

Abstract

Techniques are provided for adaptive resource allocation for workloads with initial condition setting. One method comprises obtaining a dataset comprising data from previous executions of a workload, wherein the data comprises a plurality of different resource allocations and parameterizations of the workload; determining an initial allocation of an amount of a resource for the workload based on a regression model characterizing a behavior of the workload, the data, a predefined service metric and a characterization of a target infrastructure; and initiating an application of the determined initial allocation of the amount of the resource for the workload. A performance of one or more of the plurality of workloads can be evaluated based on a percentage of time within a predefined error range. The regression model can be updated and/or replaced over time with new data for additional executions of the at least one workload.

First claim

What is claimed is:

1. A method, comprising:
obtaining a dataset comprising data from previous executions of at least one workload of a plurality of workloads, wherein the data comprises a plurality of different resource allocations for the at least one workload and a plurality of different parameterizations of one or more parameters that configure the at least one workload;
determining an initial allocation of an amount of at least one resource to be allocated to the at least one workload by applying, to a regression model characterizing a behavior of the at least one workload, (i) the data comprising at least some of the plurality of different resource allocations for the previous executions of the at least one workload and at least some of the plurality of different parameterizations of one or more parameters that configured the previous executions of the at least one workload, (ii) at least one predefined service metric for the at least one workload and (iii) a characterization of a target infrastructure where the at least one workload will execute;
initiating an application of the determined initial allocation of the amount of the at least one resource to be allocated to the at least one workload; and
updating a complexity of the regression model with at least one additional parameter using new data for one or more additional executions of the at least one workload;
wherein the method is performed by at least one processing device comprising a processor coupled to a memory.

2. The method of claim 1 , wherein the different resource allocations comprise one or more of a different number of processing cores in a computer processor, a number of different processing cores in a graphics processing unit, a different amount of memory and a different amount of network bandwidth.

3. The method of claim 1 , wherein the determining of the initial allocation of the amount of the at least one resource for the at least one workload is performed substantially in parallel with an execution of the plurality of workloads.

4. The method of claim 1 , further comprising evaluating a performance of one or more of the plurality of workloads based on a percentage of time within a predefined error range.

5. The method of claim 1 , further comprising determining an adjustment to the initial allocation of the at least one resource for the at least one workload based at least in part on one or more of (i) a dynamic system model based on a relation between the amount of the at least one resource for the plurality of workloads and the at least one predefined service metric, (ii) an interference effect of one or more additional workloads of the plurality of workloads on the at least one workload, and (iii) a difference between an instantaneous value of the at least one predefined service metric and a target value for the at least one predefined service metric.

6. The method of claim 1 , further comprising replacing the regression model over time with a different model using the new data for the one or more additional executions of the at least one workload.

7. The method of claim 1 , further comprising a plurality of the regression models, and wherein an accuracy of each of the plurality of the regression models is evaluated over time to identify a most fitting model.

8. The method of claim 1 , further comprising a plurality of the regression models, wherein an accuracy of each of the plurality of the regression models is evaluated over time and wherein at least one of the regression models is retrained when a predefined model degradation standard is violated.

9. A computer program product, comprising a tangible non-transitory machine-readable storage medium having encoded therein executable code of one or more software programs, wherein the one or more software programs when executed by at least one processing device perform the following steps:
obtaining a dataset comprising data from previous executions of at least one workload of a plurality of workloads, wherein the data comprises a plurality of different resource allocations for the at least one workload and a plurality of different parameterizations of one or more parameters that configure the at least one workload;
determining an initial allocation of an amount of at least one resource to be allocated to the at least one workload by applying, to a regression model characterizing a behavior of the at least one workload, (i) the data comprising at least some of the plurality of different resource allocations for the previous executions of the at least one workload and at least some of the plurality of different parameterizations of one or more parameters that configured the previous executions of the at least one workload, (ii) at least one predefined service metric for the at least one workload and (iii) a characterization of a target infrastructure where the at least one workload will execute;
initiating an application of the determined initial allocation of the amount of the at least one resource to be allocated to the at least one workload; and
updating a complexity of the regression model with at least one additional parameter using new data for one or more additional executions of the at least one workload.

10. The computer program product of claim 9 , further comprising evaluating a performance of one or more of the plurality of workloads based on a percentage of time within a predefined error range.

11. The computer program product of claim 9 , further comprising determining an adjustment to the initial allocation of the at least one resource for the at least one workload based at least in part on one or more of (i) a dynamic system model based on a relation between the amount of the at least one resource for the plurality of workloads and the at least one predefined service metric, (ii) an interference effect of one or more additional workloads of the plurality of workloads on the at least one workload, and (iii) a difference between an instantaneous value of the at least one predefined service metric and a target value for the at least one predefined service metric.

12. The computer program product of claim 9 , further comprising replacing the regression model over time with a different model using the new data for the one or more additional executions of the at least one workload.

13. The computer program product of claim 9 , further comprising a plurality of the regression models, and wherein an accuracy of each of the plurality of the regression models is evaluated over time to identify a most fitting model.

14. The computer program product of claim 9 , further comprising a plurality of the regression models, wherein an accuracy of each of the plurality of the regression models is evaluated over time and wherein at least one of the regression models is retrained when a predefined model degradation standard is violated.

15. An apparatus, comprising:
a memory; and
at least one processing device, coupled to the memory, operative to implement the following steps:
obtaining a dataset comprising data from previous executions of at least one workload of a plurality of workloads, wherein the data comprises a plurality of different resource allocations for the at least one workload and a plurality of different parameterizations of one or more parameters that configure the at least one workload;
determining an initial allocation of an amount of at least one resource to be allocated to the at least one workload by applying, to a regression model characterizing a behavior of the at least one workload, (i) the data comprising at least some of the plurality of different resource allocations fo

Assignees

• Emc Ip Holding Co Llc

Inventors

• Calmon Tiago Salviano
• Gottin Vinícius Michel
• Sousa Eduardo Vera

Classifications

• G06F9/5005 Primary

  to service a request · CPC title

• G06F9/5027 Primary

  the resource being a machine, e.g. CPUs, Servers, Terminals · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/063

  using electronic means · CPC title

• G06F9/5016

  the resource being the memory · CPC title