Methods, systems and computer readable media for intelligent fetching of data storage device commands from submission queues

• US 10,642,500 B2
• Filed: 02/28/2018
• Issued: 05/05/2020
• Est. Priority Date: 09/28/2015
• Status: Active Grant

• Current Assignee: SanDisk Technologies LLC (Western Digital Corporation)
• Original Assignee: SanDisk Technologies LLC (Western Digital Corporation)
• Inventors: Benisty, Shay
• Primary Examiner(s): Thai, Tuan V
• Application Number:
  US15/908,710
• Publication Number:
  US 20180188975A1
• Time in Patent Office: 797 Days
• Field of Search: 711100, 711103, 711154, 711200
• CPC Class Codes: G06F 3/0611 in relation to response time
• CPC Class Codes: G06F 3/0659 Command handling arrangements, e.g. command buffers, queues, command scheduling
• CPC Class Codes: G06F 3/0679 Non-volatile semiconductor memory device, e.g. flash memory, one time programmable memory [OTP]
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Systems and methods for intelligent fetching of data storage device commands from submission queues are provided. One such method involves fetching commands from one or more submission queues, monitoring characteristics of the commands including a command type, predicting a next command based on the monitored command characteristics, monitoring a resource state of a data storage device, selecting a submission queue based on the predicted next command and the resource state, fetching a command from the selected submission queue, and providing the command from the selected submission queue to command processing logic.

Claims

1. A method for use by a data storage controller, comprising:

fetching commands from one or more submission queues;

monitoring characteristics of the commands including a command type;

predicting, using a machine learning algorithm, a next command based on the monitored command characteristics;

monitoring a resource state of a data storage device;

selecting a submission queue based on the predicted next command and the resource state;

fetching a command from the selected submission queue; and

providing the command from the selected submission queue to command processing logic.

2. The method of claim 1, wherein the predicting the next command based on the monitored command characteristics comprises predicting a command type of the next command based on the monitored command characteristics.

3. The method of claim 2, wherein the predicting the next command based on the monitored command characteristics further comprises predicting a command attribute of the next command based on the monitored command characteristics.

4. The method of claim 1, wherein the monitored command characteristics further comprise a characteristic selected from the group consisting of a command size, a logical block address (LBA) range, a write stream identifier, a force unit access characteristic, a dataset management characteristic, an end to end protection characteristic, an association with dies of the data storage device characteristic, and combinations thereof.

5. The method of claim 1, wherein the machine learning algorithm comprises an algorithm selected from the group consisting of a logistic regression algorithm, a support vector machine algorithm, a mixture model algorithm, a Hidden Markov Model algorithm, a neural network algorithm, and combinations thereof.

6. The method of claim 1, further comprising training the machine learning algorithm before the data storage device is released to a customer.

7. The method of claim 6, further comprising training the machine learning algorithm after the data storage device is released to the customer.

8. The method of claim 1, wherein the machine learning algorithm is configured to detect a data characteristic selected from the group consisting of a pattern and a command sequence.

9. The method of claim 1, wherein the predicting the next command based on the monitored command characteristics comprises predicting, for each of the one or more submission queues, the next command based on the monitored command characteristics for the respective submission queue.

10. The method of claim 1, wherein the predicting the next command based on the monitored command characteristics comprises predicting, for each of the one or more submission queues, the next command based on the monitored command characteristics for the respective submission queue and at least one other submission queue of the one or more submission queues.

11. The method of claim 1:

wherein the selecting the submission queue based on the predicted next command and the resource state comprises selecting the submission queue having a next command that is predicted to be a write command comprising a preselected size; and

wherein the method further comprises executing the write command, wherein a partially full write cache becomes full upon the execution of the write command, thereby causing the data storage device to empty the write cache.

12. The method of claim 1:

wherein the selecting the submission queue based on the predicted next command and the resource state comprises selecting the submission queue having a next command that is predicted to be a write command and comprising a write data; and

wherein the method further comprises executing the write command such that the write data is stored to a memory location that represents a hole in a group of data relocated for garbage collection, thereby filling the hole.

13. The method of claim 1:

wherein the selecting the submission queue based on the predicted next command and the resource state comprises selecting the submission queue having a next command that is predicted to be a read command; and

wherein the method further comprises executing the read command as part of a burst read at the data storage device.

14. The method of claim 1:

wherein the selecting the submission queue based on the predicted next command and the resource state comprises selecting, if the resource state is indicative of a non-busy state, the submission queue using an algorithm selected from the group consisting of a round robin algorithm and a weighted round robin algorithm.

15. A data storage controller for use with a data storage device, comprising:

a memory;

a processing circuitry coupled to the memory and configured to;

fetch commands from one or more submission queues;

monitor characteristics of the commands including a command type;

predict, using a machine learning algorithm, a next command based on the monitored command characteristics;

monitor a resource state of a data storage device;

select a submission queue based on the predicted next command and the resource state;

fetch a command from the selected submission queue; and

provide the command from the selected submission queue to command processing logic.

16. The data storage controller of claim 15, wherein the processing circuitry is further configured to predict a command type of the next command based on the monitored command characteristics.

17. The data storage controller of claim 16, wherein the processing circuitry is further configured to predict a command attribute of the next command based on the monitored command characteristics.

18. The data storage controller of claim 15, wherein the monitored command characteristics further comprise a characteristic selected from the group consisting of a command size, a logical block address (LBA) range, a write stream identifier, a force unit access characteristic, a dataset management characteristic, an end to end protection characteristic, an association with dies of the data storage device characteristic, and combinations thereof.

19. The data storage controller of claim 15, wherein the machine learning algorithm comprises an algorithm selected from the group consisting of a logistic regression algorithm, a support vector machine algorithm, a mixture model algorithm, a Hidden Markov Model algorithm, a neural network algorithm, and combinations thereof.

20. The data storage controller of claim 15, wherein the processing circuitry is further configured to train the machine learning algorithm before the data storage device is released to a customer.

21. The data storage controller of claim 20, wherein the processing circuitry is further configured to train the machine learning algorithm after the data storage device is released to the customer.

22. The data storage controller of claim 15, wherein the machine learning algorithm is configured to detect a data characteristic selected from the group consisting of a pattern and a command sequence.

23. The data storage controller of claim 15, wherein the processing circuitry is further configured to predict, for each of the one or more submission queues, the next command based on the monitored command characteristics for the respective submission queue.

24. The data storage controller of claim 15, wherein the processing circuitry is further configured to predict, for each of the one or more submission queues, the next command based on the monitored command characteristics for the respective submission queue and at least one other submission queue of the one or more submission queues.

25. The data storage controller of claim 15, wherein the processing circuitry is further configured to:

select the submission queue having a next command that is predicted to be a write command comprising a preselected size; and

execute the write command, wherein a partially full write cache becomes full upon the execution of the write command, thereby causing the data storage device to empty the write cache.

26. The data storage controller of claim 15, wherein the processing circuitry is further configured to:

select the submission queue having a next command that is predicted to be a write command comprising a write data; and

execute the write command such that the write data is stored to a memory location that represents a hole in a group of data relocated for garbage collection, thereby filling the hole.

27. The data storage controller of claim 15, wherein the processing circuitry is further configured to:

select the submission queue having a next command that is predicted to be a read command; and

execute the read command as part of a burst read at the data storage device.

28. The data storage controller of claim 15, wherein the processing circuitry is further configured to:

select, if the resource state is indicative of a non-busy state, the submission queue using an algorithm selected from the group consisting of a round robin algorithm and a weighted round robin algorithm.

29. A data storage controller for use with a data storage device, comprising:

means for fetching commands from one or more submission queues;

means for monitoring characteristics of the commands including a command type;

means for predicting, using a machine learning algorithm, a next command based on the monitored command characteristics;

means for monitoring a resource state of a data storage device;

means for selecting a submission queue based on the predicted next command and the resource state;

means for fetching a command from the selected submission queue; and

means for providing the command from the selected submission queue to command processing logic.

30. The data storage controller of claim 29, wherein the means for selecting the submission queue based on the predicted next command and the resource state comprises:

means for selecting, if the resource state is indicative of a non-busy state, the submission queue using an algorithm selected from the group consisting of a round robin algorithm and a weighted round robin algorithm.

31. The data storage controller of claim 29, wherein the machine learning algorithm comprises an algorithm selected from the group consisting of a logistic regression algorithm, a support vector machine algorithm, a mixture model algorithm, a Hidden Markov Model algorithm, a neural network algorithm, and combinations thereof.