Write back coordination node for cache latency correction

• US 9,684,594 B2
• Filed: 07/16/2014
• Issued: 06/20/2017
• Est. Priority Date: 07/16/2014
• Status: Active Grant

• Current Assignee: ClearSky Data Co.
• Original Assignee: ClearSky Data, Inc. (Amazon.com, Inc.)
• Inventors: Suman, Daniel, Lacy, Janice Ann, Vekiarides, Lazarus
• Primary Examiner(s): Bragdon, Reginald
• Assistant Examiner(s): VO, THANH DUC
• Application Number:
  US14/332,469
• Publication Number:
  US 20160019146A1
• Time in Patent Office: 1,070 Days
• CPC Class Codes: G06F 12/0806 Multiuser, multiprocessor or multiprocessing cache systems
• CPC Class Codes: G06F 12/0811 with multilevel cache hierarchies
• CPC Class Codes: G06F 12/0897 with two or more cache hierarchy levels with multilevel cache hierarchies G06F12/0811
• CPC Class Codes: G06F 12/10 Address translation
• CPC Class Codes: G06F 2212/283 Plural cache memories
• CPC Class Codes: G06F 2212/62 Details of cache specific to multiprocessor cache arrangements

Abstract

A coordinating node acts as a write back cache, isolating local cache storage endpoints from latencies associated with accessing geographically remote cloud cache and storage resources.

Claims

1. A coordinator node apparatus for operating on data blocks comprising:

a first interface, connected to a Level 2 (L2) cache storage endpoint over a first network connection;

a second interface, connected to cloud storage over a second network connection, the first and second network connections further such that a physical distance between the coordinator node and the L2 cache storage endpoint is shorter than a physical distance between the coordinator node and the cloud storage;

a local Level 3 (L3) cache storage; and

a processor for responding to a block write request from the L2 cache storage endpoint as a write back cache, by acknowledging the block write request to the L2 cache storage endpoint even if a data block associated with the write request is not presently stored in the local L3 cache storage;

storing the data block and metadata associated with the block write request in the local L3 cache storage, before forwarding the data block and metadata associated with the block write request to the cloud storage;

maintaining a global logical block address (LBA) metadata map; and

distributing the global LBA metadata map to the L2 cache storage endpoint and to the cloud storage.

2. The apparatus of claim 1 wherein the first network connection to the L2 storage endpoint is a local area network connection or a metropolitan area network connection.

3. The apparatus of claim 1 wherein the second network connection to the cloud storage is a wide area network connection.

4. The apparatus of claim 3 wherein the cloud storage is provided by a service accessible over the wide area network connection.

5. The apparatus of claim 1 wherein a variability in latency associated with the first network connection is one-third or less of a variability in latency associated with the second network connection.

6. The apparatus of claim 1 wherein the processor is further for assigning a signature to a data block, and maintaining, in the global LBA metadata map, metadata that specifies (a) a mapping between at least one of the LBAs and the signature, and (b) a mapping between the signature and a logical address for the data block.

7. The apparatus of claim 1 wherein the L2 storage endpoint comprises two or more devices each storing a portion of the LBAs assigned by the global LBA metadata map.

8. The apparatus of claim 1 wherein the processor further deduplicates the data block in the local L3 cache storage before forwarding the data block to the cloud storage.

9. The apparatus of claim 1 wherein the processor further, upon receiving a read request from the L2 cache storage endpoint, searches the local L3 cache storage for a data block associated with the read request that was stored from a prior write request before requesting the data block from the cloud storage.

10. The apparatus of claim 1 wherein the L2 cache storage endpoint further operates as a write-through cache for access requests received from Level 1 (L1) storage devices.

11. The apparatus of claim 10 wherein acknowledging the write request to the L2 cache storage endpoint also isolates the L1 storage devices from a latency associated with storing the data associated with the write request to the remote cloud storage.

12. The apparatus of claim 1 wherein the processor further receives updates to the global LBA metadata map from the L2 cache storage endpoint.

13. A method comprising:

accessing a storage device providing a Level 3 (L3) cache;

connecting to a Level 2 (L2) cache storage endpoint over a first network connection;

connecting to remote cloud storage over a second network connection;

the first and second network connections such that a physical distance between the L3 cache and the L2 cache storage endpoint is shorter than a physical distance from the L3 cache to the remote cloud storage;

responding to a block write request from the L2 cache storage endpoint as a write back cache, by acknowledging the block write request to the L2 cache storage endpoint even if a data block associated with the write request is not presently stored in the local L3 cache;

storing the data block associated with the write request and block-related metadata in the local L3 cache before forwarding the data block associated with the write request and the block-related metadata to the remote cloud storage;

maintaining a global logical block address (LBA) metadata map; and

distributing the global LBA metadata map to the L2 cache storage endpoint and to the remote cloud storage.

14. The method of claim 13 wherein the first network connection is a local area network connection or a metropolitan area network connection.

15. The method of claim 13 wherein the second network connection is a wide area network connection.

16. The method of claim 15 wherein the remote cloud storage is provided by a service accessible over a wide area network connection.

17. The method of claim 13 wherein a latency associated with the first network connection varies over a time that is one-third or less of a time over which a latency associated with the second network connection varies.

18. The method of claim 13 further comprising steps of:

assigning a signature to the data block, and

maintaining in the global LBA metadata map, a mapping between at least one of the LBAs and the signature, and a mapping between the signature and a logical address for the data block.

19. The method of claim 13 wherein the L2 storage endpoint comprises two or more devices each such device further storing a portion of the LBAs assigned by the global LBA metadata map.

20. The method of claim 13 further comprising:

deduplicating the data in the local L3 cache before forwarding the data block associated with the write request to the remote cloud storage.

21. The method of claim 13 further comprising:

upon receiving a read request from the L2 cache storage endpoint, searching the local L3 cache for a data block associated with the read request that may be stored from a prior write request before requesting the data block from the cloud storage.

22. The method of claim 13 further comprising the step of:

at the L2 cache storage endpoint, responding to write requests received from a level 1 (L1) storage device as a write-through cache.

23. The method of claim 22 wherein acknowledging the write request to the L2 cache storage endpoint also isolates the L1 storage devices from a latency associated with storing the data associated with the write request to the remote cloud storage.

24. The method of claim 13 additionally comprising:

in response to the L2 cache storage endpoint satisfying a write request, receiving updates to a portion of the global LBA metadata map from the L2 cache storage endpoint.