Data management, reduction and sampling schemes for storage device failure

Patent metadata

Field	Value
Publication number	US-11669754-B2
Application number	US-202016872194-A
Country	US
Kind code	B2
Filing date	May 11, 2020
Priority date	Feb 25, 2020
Publication date	Jun 6, 2023
Grant date	Jun 6, 2023

Abstract

In a method for training a machine learning model, the method includes: segmenting, by a processor, a dataset from a database into one or more datasets based on time period windows; assigning, by the processor, one or more weighted values to the one or more datasets according to the time period windows of the one or more datasets; generating, by the processor, a training dataset from the one or more datasets according to the one or more weighted values; and training, by the processor, the machine learning model using the training dataset.

First claim

What is claimed is:

1. A method for training a machine learning model, the method comprising:
segmenting, by a processor, a dataset from a database into one or more datasets based on time period windows;
assigning, by the processor, one or more weighted values to the one or more datasets according to the time period windows of the one or more datasets;
generating, by the processor, a training dataset from the one or more datasets, wherein an amount of data generated from the one or more datasets is based on the one or more weighted values; and
training, by the processor, the machine learning model using the training dataset.

2. The method according to claim 1 , wherein the machine learning model comprises a solid-state drive (SSD) failure prediction model.

3. The method according to claim 1 , wherein a most recent dataset from the one or more datasets is assigned a first weighted value and a least recent dataset from the one or more datasets is assigned a second weighted value, wherein the first weighted value is greater than the second weighted value.

4. The method according to claim 3 , wherein the one or more weighted values decrease by a set amount from the first weighted value to the second weighted value.

5. The method according to claim 1 , the method further comprising:
identifying, by the processor, anomaly data in the dataset;
retrieving, by the processor, the anomaly data in the dataset; and
adding, by the processor, the anomaly data to the training dataset.

6. The method according to claim 5 , wherein the anomaly data comprises SSD failure data.

7. The method according to claim 5 , wherein the anomaly data is identified using a rule based method.

8. The method according to claim 5 , wherein the anomaly data is identified using a cluster based method.

9. The method according to claim 1 , the method further comprising
generating, by the processor, anomaly data; and
adding, by the processor, the generated anomaly data to the training dataset.

10. A data system comprising:
a database;
a processor coupled to the database; and
a memory coupled to the processor, wherein the memory stores instructions that, when executed by the processor, cause the processor to:
segment a dataset from the database into one or more datasets based on time period windows;
assign one or more weighted values to the one or more datasets according to the time period windows of the one or more datasets;
generate a training dataset from the one or more datasets, wherein an amount of data generated from the one or more datasets is based on the one or more weighted values; and
train a machine learning model using the training dataset.

11. The data system according to claim 10 , wherein the machine learning model comprises a solid-state drive (SSD) failure prediction model.

12. The data system according to claim 10 , wherein a most recent dataset from the one or more datasets is assigned a first weighted value and a least recent dataset from the one or more datasets is assigned a second weighted value, wherein the first weighted value is greater than the second weighted value.

13. The data system according to claim 12 , wherein the one or more weighted values decrease by a set amount from the first weighted value to the second weighted value.

14. The data system according to claim 10 , wherein the processor is further configured to:
identify anomaly data in the dataset;
retrieve the anomaly data in the dataset; and
add the anomaly data to the training dataset.

15. The data system according to claim 14 , wherein the anomaly data comprises SSD failure data.

16. The data system according to claim 14 , wherein the anomaly data is identified using a rule based method.

17. The data system according to claim 14 , wherein the anomaly data is identified using a cluster based method.

18. The data system according to claim 10 , wherein the processor is further configured to:
generate anomaly data; and
add the generated anomaly data to the training dataset.

19. A method for training a machine learning model, the method comprising:
identifying, by a processor, anomaly data in a dataset from a database;
generating, by the processor, additional anomaly data;
adding, by the processor, the generated anomaly data to the dataset;
identifying, by the processor, a training dataset from the dataset;
retrieving, by the processor, the training dataset from dataset; and
training, by the processor, the machine learning model using the training dataset.

20. The method according to claim 19 , wherein the machine learning model comprises a solid-state (SSD) failure prediction model.

Assignees

• Samsung Electronics Co Ltd

Inventors

• Elyasi Nima
• Sinha Vikas
• Zheng Qinling
• Choi Changho

Classifications

• G06F11/008

  Reliability or availability analysis · CPC title

• G06F11/3037

  where the computing system component is a memory, e.g. virtual memory, cache (accessing, addressing or allocating within memory systems or architectures G06F12/00; checking stores for correct operation G11C29/00) · CPC title

• G06F18/2135

  based on approximation criteria, e.g. principal component analysis · CPC title

• G06F3/0679

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

• G06F3/0619

  in relation to data integrity, e.g. data losses, bit errors · CPC title