Finding the optimum time to prestart container running functions using events

Patent metadata

Field	Value
Publication number	US-11675634-B2
Application number	US-202016887660-A
Country	US
Kind code	B2
Filing date	May 29, 2020
Priority date	May 29, 2020
Publication date	Jun 13, 2023
Grant date	Jun 13, 2023

Abstract

A method includes identifying a first event that has been at least partly performed, wherein the first event comprises an element of a sequence of events, and the first event comprises performance of a first computing function, predicting a second event expected to occur next in the sequence after completion of the first event, and the second event comprises performance of a second computing function, predicting a start time of the second event, based on information about the second event, identifying a particular container capable of implementing the second computing function associated with the second event, predicting a start time for start-up of the container, starting up the container, and completing start-up of the container prior to receipt of a request for the second computing function to be performed by the container, wherein the container is ready to perform the second computing function immediately after start-up has been completed.

First claim

What is claimed is:

1. A method, comprising:
identifying a first event including a sequence of computing functions including a first computing function, wherein starting execution of a first container for the first computing function triggers a second event of a second container capable of implementing a second computing function, which is different from the first computing function, without a separate triggering event for the second event;
predicting a container start-up time of the second container for the second computing function after completion of execution of the first container;
based on information about the second event, creating the second container that is capable of implementing the second computing function;
starting up the second container at the predicted container start-up time; and
executing the second container for the second computing function immediately after start-up has been completed.

2. The method as recited in claim 1 , wherein the second computing function is called by the first computing function.

3. The method as recited in claim 1 , wherein prediction of the container start-up time of the second container is based on information received from an event generator, and the information comprises information about the first computing function and one or more parameters relating the first computing function to the second computing function.

4. The method as recited in claim 1 , wherein the second container is identified based on a policy.

5. The method as recited in claim 1 , wherein predicting the container start-up time of the second container comprises determining a probability that the second event will occur.

6. The method as recited in claim 1 , wherein the container start-up time of the second container is predicted by a Machine Learning (ML) process using historical information about one or more other events that have already occurred.

7. The method as recited in claim 1 , wherein the container start-up time of the second container is predicted using a neural net predictor.

8. The method as recited in claim 1 , wherein the method is performed in a serverless platform.

9. The method as recited in claim 1 , wherein a time gap between the first event and the second event is predicted using historical information about one or more events that have already occurred.

10. The method as recited in claim 1 , wherein a plurality of parameters are used to predict a container start-up time for another container, and the plurality of parameters comprise a function start-up time, an event queue, and a function pool size.

11. A non-transitory storage medium having stored therein instructions that are executable by one or more hardware processors to perform operations comprising:
identifying a first event including a sequence of computing functions including a first computing function, wherein starting execution of a first container for the first computing function triggers a second event of a second container capable of implementing a second computing function, which is different from the first computing function, without a separate triggering event for the second event;
predicting a container start-up time of the second container for the second computing function after completion of execution of the first container;
based on information about the second event, creating the second container that is capable of implementing the second computing function;
starting up the second container at the predicted start-up time; and
executing the second container for the second computing function immediately after start-up has been completed.

12. The non-transitory storage medium as recited in claim 11 , wherein the second computing function is called by the first computing function.

13. The non-transitory storage medium as recited in claim 11 , wherein prediction of the container start-up time of the second container is based on information received from an event generator, and the information comprises information about the first computing function and one or more parameters relating the first computing function to the second computing function.

14. The non-transitory storage medium as recited in claim 11 , wherein the second container is identified based on a policy.

15. The non-transitory storage medium as recited in claim 11 , wherein predicting the container start-up time of the second container comprises determining a probability that the second event will occur.

16. The non-transitory storage medium as recited in claim 11 , wherein the container start-up time of the second container is predicted by a Machine Learning (ML) process using historical information about one or more other events that have already occurred.

17. The non-transitory storage medium as recited in claim 11 , wherein the container start-up time of the second container is predicted using a neural net predictor.

18. The non-transitory storage medium as recited in claim 11 , wherein the operations are performed in a serverless platform.

19. The non-transitory storage medium as recited in claim 11 , wherein a time gap between the first event and the second event is predicted using historical information about one or more events that have already occurred.

20. The non-transitory storage medium as recited in claim 11 , wherein a plurality of parameters are used to predict a container start-up time for another container, and the plurality of parameters comprise a function start-up time, an event queue, and a function pool size.

Assignees

• Emc Ip Holding Co Llc

Inventors

• Fong Victor
• He Xuebin
• Roy Prasenjit
• Padinharepatt Nisanth Mathilakath

Classifications

• G06F9/4843 Primary

  by program, e.g. task dispatcher, supervisor, operating system · CPC title

• G06F9/546

  Message passing systems or structures, e.g. queues · CPC title

• G06F2009/45575

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

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G06F9/542 Primary

  Event management; Broadcasting; Multicasting; Notifications · CPC title