Data transfer command latency of a host device

• US 10,642,764 B1
• Filed: 03/01/2019
• Issued: 05/05/2020
• Est. Priority Date: 03/01/2019
• Status: Active Grant

• Current Assignee: Sandisk Technologies Incorporated (Western Digital Corporation)
• Original Assignee: Western Digital Technologies Incorporated (Western Digital Corporation)
• Inventors: Gerhart, Darin Edward, Ortmeier, Nicholas Edward, Lappi, Cory, Walker, William Jared
• Primary Examiner(s): Tsai, Henry
• Assistant Examiner(s): Bartels, Christopher A
• Application Number:
  US16/290,705
• Time in Patent Office: 431 Days
• Field of Search: None
• CPC Class Codes: G06F 12/0246 in block erasable memory, e.g. flash memory
• CPC Class Codes: G06F 12/10 Address translation
• CPC Class Codes: G06F 13/20 for access to input/output bus
• CPC Class Codes: G06F 13/4282 on a serial bus, e.g. I2C bus, SPI bus on daisy chain buses G06F13/4247
• CPC Class Codes: G06F 2212/657 Virtual address space management
• CPC Class Codes: G06F 2212/7201 Logical to physical mapping or translation of blocks or pages
• CPC Class Codes: G06F 2213/0036 Small computer system interface [SCSI]

Abstract

An input/output device identifies a workload type for a connected host device. The input/output device applies settings for the workload type to affect one or more of a flash translation layer queue depth of the input/output device, a host device data transfer scheduler of the input/output device, and a command status host notification timing by the input/output device to the host device.

Claims

1. A method of operating an input/output device, the method comprising:

the input/output device interacting with a host device workload execution profiler to achieve a workload type lock;

the input/output device correlating the workload type lock to a workload type;

comparing a host device workload threshold with one or more host perfection scores;

wherein the host perfection scores are calculated for each data input or data output initiator of the host device using a ranking mechanism;

wherein each host perfection score is determined by;

(W1*A1)+(W2*A2)+ . . . +(Wn*An)

where A1, A2, through An are host measurement attributes and W1, W2, through Wn are relative weight factors applied to each of the host measurement attributes and wherein the sum of the relative weight factors is 100%; and

the input/output device altering settings for the workload type to affect one of a flash translation layer queue depth, a host device data transfer scheduler, and a command status host notification timing to a host device, based on the host perfection scores.

2. The method of claim 1, further comprising measuring a difference between an actual performance measure for the workload type and an optimal performance for the workload type.

3. The method of claim 1, further comprising:

evaluating and scoring changes to performance of the host device related to changes to prior control settings for the input/output device; and

adjusting weights applied for selection of future control settings for the input/output device based on the scoring of the changes.

4. The method of claim 1, further comprising:

making a first determination whether to alter a queue depth of a flash translation layer of the input/output device based on the host perfection scores.

5. The method of claim 4, further comprising:

making a second determination, subsequent to the first determination, whether delays for host data transfers should be altered based on the host perfection scores.

6. The method of claim 5, further comprising:

making a third determination, subsequent to the second determination, whether command status responses to the host device should be delayed based on the host perfection scores.

7. A solid state drive comprising:

a read data request and write data request interface to a host device; and

a controller configured to;

identify a workload type for the host device based at least in part on activity on the interface to the host device;

compare a host device workload threshold with one or more host perfection scores;

calculate the host perfection scores for each data input or data output initiator of the host device using a ranking mechanism;

wherein the host perfection scores are determined by;

(W1*A1)+(W2*A2)+ . . . +(Wn*An)

where A1, A2, through An are host measurement attributes and W1, W2, through Wn are relative weight factors applied to each host measurement attribute and wherein the sum of the relative weight factors is 100%; and

alter settings for the workload type to first affect a flash translation layer queue depth of the solid state drive, to second affect a host device data transfer scheduler of the solid state drive, and to third effect a command status host notification timing by the solid state drive to the host device, based on the host perfection scores.

8. The solid state drive of claim 7, further comprising:

the controller configured to apply changes to performance of the host device related to changes to the settings to adjust machine learning weights applied for selection of future control settings for the solid state drive for the workload type.

9. The solid state drive of claim 7, further comprising:

a mode page comprising workload descriptors for the workload type; and

the controller configured to utilize the mode page to determine host performance attributes for the workload type.

10. The solid state drive of claim 9, wherein the host performance attributes include one or more of command turnaround time, host data turnaround time, command-to-command turnaround time, open connection time, and Serial Attached SCSI credit turnaround time.

11. An input/output device for a host device, the input/output device comprising:

a host device interface;

logic to operate a host device workload execution profiler in conjunction with the host device interface to obtain a workload type lock;

logic to correlate the workload type lock to a workload type;

logic to compare a host device workload threshold with one or more host perfection scores;

wherein the host perfection scores are calculated for each data input or data output initiator of the host device using a ranking mechanism;

wherein the host perfection score is determined by;

(W1*A1)+(W2*A2)+ . . . +(Wn*An)

where A1, A2, through An are host measurement attributes and W1, W2, through Wn are relative weight factors applied to each host measurement attribute and wherein the sum of the relative weight factors is 100%; and

logic to alter settings for the workload type to affect one or more of a flash translation layer queue depth of the input/output device, a host device data transfer scheduler of the input/output device, and a command status host notification timing by the input/output device to the host device, based on the host perfection scores.

12. The input/output device of claim 11, further comprising logic to determine a difference between an actual performance measure for the workload type and an optimal performance for the workload type.

13. The input/output device of claim 11, further comprising:

logic to determine scores for changes to performance of the host device related to changes to prior control settings for the input/output device; and

logic to adjust weights applied for selection of future control settings for the input/output device based on the scores.

14. The input/output device of claim 11, further comprising:

logic to make a first determination whether to alter a queue depth of a flash translation layer of the input/output device based on the host perfection scores; and

logic to make a second determination whether delays for host data transfers should be altered based on the host perfection scores, the second determination made on condition that that the queue depth is not altered.

15. The input/output device of claim 14, further comprising:

logic to make a third determination whether command status responses to the host device should be delayed based on the host perfection scores, the third determination made on condition that the delays for host data transfers are not altered.