Power down workload estimation

Patent metadata

Field	Value
Publication number	US-11853558-B2
Application number	US-202217718386-A
Country	US
Kind code	B2
Filing date	Apr 12, 2022
Priority date	Dec 30, 2021
Publication date	Dec 26, 2023
Grant date	Dec 26, 2023

Abstract

Apparatuses and methods can be related power down workload estimations using artificial neural networks. Workload estimation can include predicting a duration of a subsequent power down event of the memory device. A quantity of maintenance operations to be performed on the memory device, may be predicted based on the predicted duration of the subsequent power down event, when the memory device is powered on after the subsequent power down event using an artificial neural network. The quantity of maintenance operations may be performed on the memory device prior to the subsequent power down event of the memory device.

First claim

What is claimed is:

1. An apparatus, comprising:
a memory device;
a processing device coupled to the memory device and configured to:
predict a duration of a subsequent power down event of the memory device;
predict a quantity of maintenance operations to be performed on the memory device based on the predicted duration of the subsequent power down event when the memory device is powered on after the subsequent power down event using an artificial neural network (ANN); and
perform the quantity of maintenance operations on the memory device prior to the subsequent powering down event of the memory device.

2. The apparatus of claim 1 , wherein the processing device is further configured to compare the predicted duration of the subsequent power down event of the memory device to an actual duration of the subsequent power down event of the memory device.

3. The apparatus of claim 2 , wherein the processing device is further configured to train the ANN based on the comparison for future predictions of durations of subsequent power down events of the memory device.

4. The apparatus of claim 1 , wherein the processing device is further configured to predict the duration of the subsequent power down event of the memory device utilizing a different ANN.

5. The apparatus of claim 1 , wherein the processing device is further configured to predict the quantity of maintenance operations using the ANN comprising a long short-term memory (LSTM) network architecture.

6. The apparatus of claim 1 , wherein the processing device is further configured to train the ANN at run-time utilizing a plurality of previous power down durations.

7. The apparatus of claim 1 , wherein the ANN is trained by a manufacturer of the memory device.

8. The apparatus of claim 1 , wherein the processing device configured to perform the quantity of maintenance operations is further configured to perform wear leveling operations.

9. The apparatus of claim 8 , wherein the processing device is further configured to:
provide a logical saturation level of the memory device as an input to the ANN; and
receive a quantity of wear leveling operations to perform prior to the power down event and a predicted power down time from the ANN.

10. A method, comprising:
predicting, by a processing device coupled to a memory device, a duration of a subsequent power down event of the memory device;
predicting a quantity of maintenance operations to be performed on the memory device based on the predicted duration of the subsequent power down event when the memory device is powered on after the subsequent power down event using an artificial neural network (ANN);
performing the quantity of maintenance operations on the memory device prior to the subsequent powering down event of the memory device;
comparing the predicted duration of the subsequent power down event to an actual duration of the subsequent power down event when the memory device is powered on; and
training the ANN with the comparison of the predicted duration of the subsequent power down event to the actual duration of the subsequent power down event.

11. The method of claim 10 , wherein performing the quantity of maintenance operations further comprises performing a quantity of temperature remediation operations.

12. The method of claim 11 , wherein predicting of the quantity of maintenance operations to be performed further comprises:
providing an average temperature of the memory device and an amount of time soaking at high temperatures of the memory device to the ANN as inputs; and
receiving an amount of temperature remediation operations to perform prior to the power down event from the ANN.

13. The method of claim 10 , wherein performing the quantity of maintenance operations further comprises performing a quantity of garbage collection operations.

14. The method of claim 13 , wherein predicting of the quantity of maintenance operations to be performed further comprises:
providing a last power down time duration of the memory device, a free block pool size of the memory device, and an amount of a host traffic since the previous power down event of the memory device to the ANN as inputs; and
receiving the quantity of garbage collection operations to perform prior to the power down event from the ANN.

15. The method of claim 10 , further comprising ordering the quantity of maintenance operations to be performed so that a user experiences minimal interference with use of the memory device when powered on.

16. A system, comprising:
a memory device;
a first artificial neural network (ANN) configured to predict a duration of a subsequent power down event of the memory device;
a second ANN configured to predict a plurality of different types of maintenance operations to be performed on the memory device based on the predicted duration of the subsequent power down event when the memory device is powered on after the subsequent power down event using the first ANN; and
a processing device coupled to the first ANN and the second ANN and configured to perform the plurality of different types of maintenance operations on the memory device prior to the subsequent powering down event of the memory device.

17. The system of claim 16 , wherein the processing device configured to perform the plurality of different types of maintenance operations is further configured to perform a quantity of media scan operations.

18. The system of claim 17 , wherein the processing device is further configured to:
provide device geometry of the memory device and a time since a previous power down event of the memory device to the second ANN as an input; and
receive the quantity of media scan operations to perform prior to the power down event from the second ANN.

19. The system of claim 16 , wherein the processing device is further configured to update the second ANN utilizing weights and biases received from a cloud system.

20. The system of claim 16 , wherein the processing device, the first ANN, and the second ANN are resident on a mobile computing device.

Assignees

• Micron Technology Inc

Inventors

• Palmer David A
• Parry Jonathan S
• Basu Reshmi

Classifications

• H04H20/67

  Common-wave systems, i.e. using separate transmitters operating on substantially the same frequency · CPC title

• H04H20/103 Primary

  Transmitter-side switching · CPC title

• G06F3/0617 Primary

  in relation to availability · CPC title

• G06F3/0653

  Monitoring storage devices or systems · CPC title

• G06F3/0679

  Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP] · CPC title