Performing an atomic write operation across multiple storage devices

• US 9,471,448 B2
• Filed: 12/10/2014
• Issued: 10/18/2016
• Est. Priority Date: 12/10/2014
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Williams, Dan J., Veal, Bryan E., Foong, Annie, Trika, Sanjeev N.
• Primary Examiner(s): Lottich, Joshua P
• Application Number:
  US14/566,636
• Publication Number:
  US 20160170850A1
• Time in Patent Office: 678 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 11/1474 in transactions G06F16/20 takes precedence
• CPC Class Codes: G06F 3/0619 in relation to data integrity, e.g. data losses, bit errors
• CPC Class Codes: G06F 3/064 Management of blocks
• CPC Class Codes: G06F 3/065 Replication mechanisms
• CPC Class Codes: G06F 3/0689 Disk arrays, e.g. RAID, JBOD

Abstract

Provided are a computer program product, system, and method performing an atomic write operation across multiple storage devices. A determination is made of a plurality of storage devices on which to write data for a write operation. A tag is generated to uniquely identify the write operation. A write command is sent to each of the determined storage devices including the tag and write data to cause each of the storage devices to write the write data at the storage device. Each of the storage devices maintains a copy of a previous version of the data to be updated by the write operation. A revert command is sent with the tag to one of the storage devices to cause the storage device to restore the copy of the previous version of the write data at the storage device.

Claims

1. A computer readable storage device having computer program instructions executed in a host system in communication with a plurality of storage devices to cause operations to:

determine a plurality of storage devices on which to write data for a write operation;

generate a tag to uniquely identify the write operation;

send a write command to each of the determined storage devices including the tag and write data to cause each of the storage devices to write the write data at the storage device, wherein each of the storage devices maintains a copy of a previous version of the data to be updated by the write operation; and

send a revert command with the tag to one of the storage devices to cause the storage device to restore the copy of the previous version of the data at the storage device.

2. The computer readable storage device of claim 1, wherein the computer program instructions are further executed to:

send an open command to each of the storage devices having the tag, wherein the open command causes each of the storage devices to maintain the copy of the previous version of the data to be updated by the write operation; and

wherein the write command is sent to each of the storage devices in response to receiving acknowledgment of the open command from all of the storage devices that were sent the open command.

3. The computer readable storage device of claim 1, wherein the computer program instructions are further executed to:

in response to receiving acknowledgment for all of the storage devices that the write command completed at the storage device, send a close command to each of the storage devices, wherein the close command causes each of the storage devices to remove context of the write operation including the copy of the previous version of the data.

4. The computer readable storage device of claim 1, wherein the computer program instructions are further executed to:

detect an interruption;

determining that all the storage devices have not completed the write command; and

sending a close command to each of the storage drives in an open state to close the write operation.

5. The computer readable storage device of claim 1, wherein the computer program instructions are further executed to:

in response to detecting an interruption, determine a state of the storage devices that have not acknowledged completing the write command, wherein the revert command is sent to at least one of the storage devices determined to have completed the write command.

6. The computer readable storage device of claim 5, wherein the determine the state of the storage devices comprises sending an inquiry to each of the storage devices having received the write command that have not acknowledged the write command completed, wherein the revert command is sent to each of the storage devices that responded to the write command indicating that the write completed and to each of the storage devices that responded to the inquiry indicating that the write completed.

7. The computer readable storage device of claim 1, wherein the computer program instructions are further executed to:

cache the write data sent with the write commands;

discard the cached write data sent with one of the write commands in response to receiving acknowledgment that the write command having the write data completed;

in response to detecting an interruption, determine a state of the storage devices that have received the write command and have not acknowledged that the write completed; and

resend the write command with the cached write data to each of the storage devices having the state indicating that the write did not complete.

8. The computer readable storage device of claim 7, wherein the write commands sent to the storage devices include non-parity data sent to a plurality of the storage devices and parity data sent to at least one of the storage devices.

9. The computer readable storage device of claim 8, wherein in response to determining that the write command of the parity data did not complete, the computer program instructions are further executed to:

read the write data from each of the storage devices having the state indicating that the write command completed; and

calculate the parity data from the write data read from the at least one of the storage devices and the cached write data for storage devices having the state indicating that the write did not complete.

10. The computer readable storage device of claim 1, wherein a plurality of write commands are sent to each of the storage devices to write a portion of the write data for the storage device, wherein the computer program instructions are further executed to:

cache the portions of write data sent with the write commands;

in response to detecting an interruption, determine a state of the storage devices that have received the write commands and have not acknowledged that all the writes completed;

for each of the storage devices that have acknowledged that less than all the sent write commands completed, sending at least one read command to the storage device to read the portion of the write data;

calculating parity data from the read portion of the write data and from the cached portions of the write data sent with write commands that did not complete; and

resending the write commands with the cached portion of the write data that did not complete and for the calculated parity data to the at least one of the storage devices for which the portions of the write data are intended.

11. The computer readable storage device of claim 1, wherein the host system sends a plurality of write operations to the storage devices and for each write operation provides a tag uniquely identifying the write operation, wherein the computer program instructions are further executed to:

in response to detecting an interruption to the write operation, send an inquiry request to the storage devices for status for the tags maintained at the storage devices; and

receive, in response to the inquiry request, from each of the storage devices the tags of the write operations initiated at the storage devices and a status of the write operation for each of the tags indicating an open or written state.

12. A storage device comprising:

a controller to communicate with a host system and to;

receive, by the controller, a write command from the host system with a tag uniquely identifying a write operation and write data;

in response to the write command, write the write data to non-volatile memory in the storage device and maintain a copy of a previous version of data updated as part of the write command;

receive a revert command with the tag identifying the write operation for which the write command was received; and

in response to the revert command, restore the copy of the previous version of the data updated by the write data.

13. The storage device of claim 12, wherein the write command comprises a plurality of write commands each having a portion of write data for the write operation, wherein the write of the write data and the maintain the copy of the previous version of the data are performed for each of the write commands, wherein the controller is further to:

return acknowledgment to the write command in response to completing the writing of portions of the write data in all of the write commands; and

indicate the portion of the write data that has been written for one of the write commands as dirty.

14. The storage device of claim 12, wherein the controller is further to:

receive a close command for the write operation;

delete information on the write operation and the previous copy of the data for the write operation in response to the close command for the write operation; and

return acknowledgment to the host system of completion of the close command.

15. The storage device of claim 12, wherein the storage device comprises a solid state disk (SSD).

16. The storage device of claim 12, wherein the controller is further to:

receive an open command with the tag from the host system for the write operation;

indicate intent logging for the tag to maintain a copy of a previous version of data updated as part of the write operation identified by the tag; and

return acknowledgment of the open command in response to indicating the intent logging, wherein the write command is sent after returning acknowledgment of the open command.

17. The storage device of claim 12, wherein the controller is further to:

receive a state inquiry from the host system;

in response to receiving the state inquiry;

complete a pending write command for the write operation;

return write complete to the host system in response to completing the pending write command; and

return open state in response to determining there is no pending write command for the write operation.

18. The storage device of claim 17, wherein the complete the pending write command and return the write complete or open state to the host system are performed for each of a plurality of write operations open at the storage device.

19. A system in communication with a plurality of storage devices, comprising:

a processor; and

a computer readable storage medium having program instructions executed by the processor to;

determine a plurality of storage devices on which to write data for a write operation;

generate a tag to uniquely identify the write operation;

send a write command to each of the determined storage devices including the tag and write data to cause each of the storage devices to write the write data at the storage device, wherein each of the storage devices maintains a copy of a previous version of the data to be updated by the write operation; and

send a revert command with the tag to one of the storage devices to cause the storage device to restore the copy of the previous version of the data at the storage device.

20. The system of claim 19, wherein the program instructions are further executed to:

send an open command to each of the storage devices having the tag, wherein the open command causes each of the storage devices to maintain the copy of the previous version of the data to be updated by the write operation; and

wherein the write command is sent to each of the storage devices in response to receiving acknowledgment of the open command from all of the storage devices that were sent the open command.

21. The system of claim 19, wherein the program instructions are further executed to:

in response to detecting an interruption, determine a state of the storage devices that have not acknowledged completing the write command, wherein the revert command is sent to at least one of the storage devices determined to have completed the write command.

22. A method, comprising:

determining a plurality of storage devices on which to write data for a write operation;

generating a tag to uniquely identify the write operation;

sending a write command to each of the determined storage devices including the tag and the write data to cause each of the storage devices to write the write data at the storage device, wherein each of the storage devices maintains a copy of a previous version of the data to be updated by the write operation; and

sending a revert command with the tag to one of the storage devices to cause the storage device to restore the copy of the previous version of the data at the storage device.

23. The method of claim 22, further comprising:

sending an open command to each of the storage devices having the tag, wherein the open command causes each of the storage devices to maintain the copy of the previous version of the data to be updated by the write operation; and

wherein the write command is sent to each of the storage devices in response to receiving acknowledgment of the open command from all of the storage devices that were sent the open command.

24. The method of claim 22, further comprising:

in response to detecting an interruption, determining a state of the storage devices that have not acknowledged completing the write command, wherein the revert command is sent to at least one of the storage devices determined to have completed the write command.

25. The method of claim 22, further comprising:

caching the write data sent with the write commands;

discarding the cached write data sent with one of the write commands in response to receiving acknowledgment that the write command having the write data completed;

in response to detecting an interruption, determining a state of the storage devices that have received the write command and have not acknowledged that the write completed; and

resending the write command with the cached write data to each of the storage devices having the state indicating that the write did not complete.