SCALING PERFORMANCE FOR RAID STORAGE CONTROLLERS BY PREDICTIVELY CACHING DATA FOR HOST WRITE REQUESTS

• US 20160170639A1
• Filed: 12/15/2014
• Published: 06/16/2016
• Est. Priority Date: 12/15/2014
• Status: Active Grant

• Current Assignee: Avago Technologies International Sales Pte Limited (Broadcom, Inc.)
• Original Assignee: Avago Technologies General IP PTE Limited (Broadcom, Inc.)
• Inventors: Talamacki, Varadaraj S., Velayudhan, Vinu
• Granted Patent: US 9,471,229 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 11/1076 Parity data used in redundant arrays of independent storages, e.g. in RAID systems
• CPC Class Codes: G06F 3/061 Improving I/O performance
• CPC Class Codes: G06F 3/0659 Command handling arrangements, e.g. command buffers, queues, command scheduling
• CPC Class Codes: G06F 3/0689 Disk arrays, e.g. RAID, JBOD

Abstract

Methods and structure for predictively caching data to service host write requests. One embodiment includes a Redundant Array of Independent Disks (RAID) storage controller able to manage a RAID volume implemented on a plurality of storage devices. The storage controller includes a memory able to store a queue of write requests that are directed to the RAID volume, and an Input/Output (I/O) processor. The I/O processor is able to detect a stream of write requests in the queue that are directed to a sequence of Logical Block Addresses (LBAs) at the RAID volume, to predict, based on the stream, LBAs for new write requests, to cache data for the predicted LBAs from the storage devices to the storage controller, to receive the new write requests, and to utilize the cached data from the predicted LBAs to generate parity data for the new write requests.

Claims

1. A Redundant Array of Independent Disks (RAID) storage controller operable to manage a RAID volume implemented on a plurality of storage devices, the storage controller comprising:

a memory operable to store a queue of write requests that are directed to the RAID volume; and

an Input/Output (I/O) processor operable to detect a stream of write requests in the queue that are directed to a sequence of Logical Block Addresses (LBAs) at the RAID volume, to predict, based on the stream, LBAs for new write requests, to cache data for the predicted LBAs from the storage devices to the storage controller, to receive the new write requests, and to utilize the cached data for the predicted LBAs to generate parity data for the new write requests.

2. The storage controller of claim 1, wherein:

the I/O processor is further operable to predict LBAs for new write requests by selecting a number of LBAs to read ahead of the sequence at the RAID volume.

3. The storage controller of claim 2, wherein:

the I/O processor is further operable to generate one read operation per storage device to retrieve the number of LBAs for storage in the cache.

4. The storage controller of claim 2, wherein:

the selected number of LBAs continues the sequence of LBAs.

5. The storage controller of claim 1, wherein:

the I/O processor is further operable to detect write requests in the queue that will trigger a Full Stripe Write (FSW) operation, to detect write requests in the queue that will trigger a Read-Modify-Write (RMW) operation, and to utilize the cached data for the predicted LBAs to generate parity data only for the for the write requests that will trigger an RMW operation.

6. The storage controller of claim 1, wherein:

the I/O processor is further operable to detect that a write request in the queue is directed to a portion of a stripe of the RAID volume, and to utilize the cached data for the predicted LBAs to generate parity information for the write request instead of triggering a Read-Modify-Write (RMW) operation.

7. The storage controller of claim 1, wherein:

the I/O processor is further operable to utilize the cached data for the predicted LBAs to generate exclusive OR (XOR) data for new write requests that enter the queue.

8. A method comprising:

identifying a queue of write requests at a storage controller that are directed to a Redundant Array of Independent Disks (RAID) volume implemented on a plurality of storage devices;

detecting a stream of write requests in the queue that are directed to a sequence of Logical Block Addresses (LBAs) at the RAID volume;

predicting, based on the stream, LBAs for new write requests;

caching data for the predicted LBAs from the storage devices to the storage controller;

receiving the new write requests; and

utilizing the cached data for the predicted LBAs to generate parity data for the new write requests.

9. The method of claim 8, wherein:

predicting LBAs for new write requests comprises selecting a number of LBAs to read ahead of the sequence at the RAID volume.

10. The method of claim 9, further comprising:

generating one read operation per storage device to retrieve the number of LBAs for storage in the cache.

11. The method of claim 9, wherein:

the selected number of LBAs continues the sequence of LBAs.

12. The method of claim 8, further comprising:

detecting write requests in the queue that will trigger a Full Stripe Write (FSW) operation;

detecting write requests in the queue that will trigger a Read-Modify-Write (RMW) operation; and

utilizing the cached data for the predicted LBAs to generate parity data only for the for the write requests that will trigger an RMW operation.

13. The method of claim 8, further comprising:

detecting that a write request in the queue is directed to a portion of a stripe of the RAID volume; and

utilizing the cached data for the predicted LBAs to generate parity information for the write request instead of triggering a Read-Modify-Write (RMW) operation.

14. The method of claim 8, further comprising:

utilizing the cached data for the predicted LBAs to generate exclusive OR (XOR) data for new write requests that enter the queue.

15. A non-transitory computer readable medium embodying programmed instructions which, when executed by a processor, are operable for performing a method comprising:

identifying a queue of write requests at a storage controller that are directed to a Redundant Array of Independent Disks (RAID) volume implemented on a plurality of storage devices;

detecting a stream of write requests in the queue that are directed to a sequence of Logical Block Addresses (LBAs) at the RAID volume;

predicting, based on the stream, LBAs for new write requests;

caching data for the predicted LBAs from the storage devices to the storage controller;

receiving the new write requests; and

utilizing the cached data for the predicted LBAs to generate parity data for the new write requests.

16. The medium of claim 15, wherein:

predicting LBAs for new write requests comprises selecting a number of LBAs to read ahead of the sequence at the RAID volume.

17. The medium of claim 16, wherein the method further comprises:

generating one read operation per storage device to retrieve the number of LBAs for storage in the cache.

18. The medium of claim 16, wherein:

the selected number of LBAs continues the sequence of LBAs.

19. The medium of claim 15, wherein the method further comprises:

detecting write requests in the queue that will trigger a Full Stripe Write (FSW) operation;

detecting write requests in the queue that will trigger a Read-Modify-Write (RMW) operation; and

utilizing the cached data for the predicted LBAs to generate parity data only for the for the write requests that will trigger an RMW operation.

20. The medium of claim 15, wherein the method further comprises:

detecting that a write request in the queue is directed to a portion of a stripe of the RAID volume; and

utilizing the cached data for the predicted LBAs to generate parity information for the write request instead of triggering a Read-Modify-Write (RMW) operation.

21. A Redundant Array of Independent Disks (RAID) storage controller operable to manage a RAID volume, the storage controller comprising:

a memory operable to store host write requests that are directed to Logical Block Addresses (LBAs) at the RAID volume; and

an Input/Output (I/O) processor operable to anticipate, based on the stored host write requests, LBAs that will be accessed in the future by host write requests that have yet to be received at the storage controller, to cache data for the anticipated LBAs from persistent storage at the RAID volume to the storage controller, to receive new host write requests, and to utilize the cached data for the predicted LBAs to service the newly received host write requests.

22. The storage controller of claim 21, wherein:

the I/O processor is further operable to utilize the cached data by generating parity data for the newly received host write requests.

23. The storage controller of claim 21, wherein:

the storage controller generates Serial Attached Small Computer System Interface (SAS) Input/Output (I/O) operations to update the RAID volume with information from the newly received host write requests.

24. A method comprising:

storing, at a storage controller, host write requests that are directed to Logical Block Addresses (LBAs) of a Redundant Array of Independent Disks (RAID) volume;

anticipating, based on the stored host write requests, LBAs that will be accessed in the future by host write requests that have yet to be received at the storage controller;

caching data for the anticipated LBAs from persistent storage at the RAID volume to the storage controller;

receiving new host write requests; and

utilizing the cached data from the anticipated LBAs to service the newly received host write requests.

25. The method of claim 24, further comprising:

utilizing the cached data by generating parity data for the newly received host write requests.