Predicting time-to-finish of a workflow using deep neural network with biangular activation functions

Patent metadata

Field	Value
Publication number	US-10853718-B2
Application number	US-201816040771-A
Country	US
Kind code	B2
Filing date	Jul 20, 2018
Priority date	Jul 20, 2018
Publication date	Dec 1, 2020
Grant date	Dec 1, 2020

Abstract

Techniques are provided for predicting a time-to-finish of at least one workflow in a shared computing environment using a deep neural network with a biangular activation function. An exemplary method comprises: obtaining a specification of an executing workflow of multiple concurrent workflows in a shared computing environment, wherein the specification comprises states of past executions of the executing workflow; obtaining a trained deep neural network, wherein the trained deep neural network is trained to predict one or more future states of the executing workflow using the states of past executions and wherein the trained deep neural network employs a biangular activation function comprising multiple parameters that define a position and a slope associated with two angles of the biangular activation function for a range of input values; and estimating, using the at least one trained deep neural network, a time-to-finish of the executing workflow of the multiple concurrent workflows.

First claim

What is claimed is:

1. A method, comprising:
obtaining a specification of at least one executing workflow of a plurality of concurrent workflows in a shared computing environment, wherein the specification comprises a plurality of states of past executions of the at least one executing workflow on at least one processing device;
obtaining at least one trained deep neural network, wherein said at least one trained deep neural network is trained to predict one or more future states of the at least one executing workflow using the plurality of states of past executions and wherein said at least one trained deep neural network employs a biangular activation function comprising a substantially minimal number of parameters that define a position and a slope associated with two angles of the biangular activation function for a range of input values; and
estimating, using at least one processing device and the at least one trained deep neural network, a time-to-finish of the at least one executing workflow of the plurality of concurrent workflows.

2. The method of claim 1 , wherein the substantially minimal number of parameters comprise four parameters learned during a training phase.

3. The method of claim 1 , wherein the substantially minimal number of parameters comprise four parameters fixed during an initialization phase based on a priori information.

4. The method of claim 1 , further comprising the step of updating the at least one trained deep neural network using one or more snapshots collected during an execution of a new instance of the at least one executing workflow.

5. The method of claim 1 , wherein the time-to-finish of the at least one executing workflow is used to update an execution environment of the at least one workflow.

6. The method of claim 4 , wherein the updating of the execution environment of the at least one workflow comprises updating an allocation of resources for the at least one workflow.

7. The method of claim 1 , wherein the biangular activation function adapts locally to a level of noise in one or more input values.

8. A computer program product, comprising a 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 specification of at least one executing workflow of a plurality of concurrent workflows in a shared computing environment, wherein the specification comprises a plurality of states of past executions of the at least one executing workflow on at least one processing device;
obtaining at least one trained deep neural network, wherein said at least one trained deep neural network is trained to predict one or more future states of the at least one executing workflow using the plurality of states of past executions and wherein said at least one trained deep neural network employs a biangular activation function comprising a substantially minimal number of parameters that define a position and a slope associated with two angles of the biangular activation function for a range of input values; and
estimating, using the at least one trained deep neural network, a time-to-finish of the at least one executing workflow of the plurality of concurrent workflows.

9. The computer program product of claim 8 , wherein the substantially minimal number of parameters comprise four parameters learned during a training phase.

10. The computer program product of claim 8 , wherein the substantially minimal number of parameters comprise four parameters fixed during an initialization phase based on a priori information.

11. The computer program product of claim 8 , further comprising the step of updating the at least one trained deep neural network using one or more snapshots collected during an execution of a new instance of the at least one executing workflow.

12. The computer program product of claim 8 , wherein the time-to-finish of the at least one executing workflow is used to update an execution environment of the at least one workflow.

13. The computer program product of claim 8 , wherein the biangular activation function adapts locally to a level of noise in one or more input values.

14. An apparatus, comprising:
a memory; and
at least one processing device, coupled to the memory, operative to implement the following steps:
obtaining a specification of at least one executing workflow of a plurality of concurrent workflows in a shared computing environment, wherein the specification comprises a plurality of states of past executions of the at least one executing workflow on at least one processing device;
obtaining at least one trained deep neural network, wherein said at least one trained deep neural network is trained to predict one or more future states of the at least one executing workflow using the plurality of states of past executions and wherein said at least one trained deep neural network employs a biangular activation function comprising a substantially minimal number of parameters that define a position and a slope associated with two angles of the biangular activation function for a range of input values; and
estimating, using the at least one trained deep neural network, a time-to-finish of the at least one executing workflow of the plurality of concurrent workflows.

15. The apparatus of claim 14 , wherein the substantially minimal number of parameters comprise four parameters learned during a training phase.

16. The apparatus of claim 14 , wherein the substantially minimal number of parameters comprise four parameters fixed during an initialization phase based on a priori information.

17. The apparatus of claim 14 , further comprising the step of updating the at least one trained deep neural network using one or more snapshots collected during an execution of a new instance of the at least one executing workflow.

18. The apparatus of claim 14 , wherein the time-to-finish of the at least one executing workflow is used to update an execution environment of the at least one workflow.

19. The apparatus of claim 18 , wherein the updating of the execution environment of the at least one workflow comprises updating an allocation of resources for the at least one workflow.

20. The apparatus of claim 14 , wherein the biangular activation function adapts locally to a level of noise in one or more input values.

Assignees

• Emc Ip Holding Co Llc

Inventors

• Gottin Vinícius Michel
• Bordignon Alex Laier

Classifications

• G06N3/048

  Activation functions · CPC title

• G06N3/08 Primary

  Learning methods · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06F2009/45575

  Starting, stopping, suspending or resuming virtual machine instances · CPC title