Systems and methods for scalable object storage

• US 8,745,095 B2
• Filed: 09/21/2012
• Issued: 06/03/2014
• Est. Priority Date: 08/12/2011
• Status: Active Grant

• Current Assignee: Nexenta by DDN, Inc.
• Original Assignee: Nexenta Systems Incorporated (DataDirect Networks Incorporated)
• Inventors: Bestler, Caitlin, Aizman, Alexander
• Primary Examiner(s): Chbouki, Tarek
• Application Number:
  US13/624,593
• Publication Number:
  US 20130226978A1
• Time in Patent Office: 620 Days
• Field of Search: None
• US Class Current: 707/802
• CPC Class Codes: G06F 16/1752 based on file chunks
• CPC Class Codes: G06F 16/22 Indexing; Data structures therefor; Storage structures

Abstract

The present disclosure provides an advantageous systems and related methods for scalable object storage. One embodiment relates to a system that includes at least a manifest tracking subsystem and a chunk payload subsystem. The manifest tracking subsystem stores version manifests and encodes forward-references from the version manifests to chunks. The chunk payload subsystem stores the chunks and encodes back-references to version manifests that reference the chunk. Other embodiments, aspects and features are also disclosed.

Claims

1. A system for providing distributed storage of data objects, the system comprising:

a manifest tracking subsystem comprising a plurality of manifest server computers communicatively interconnected by a network, the manifest tracking subsystem encoding forward-references from version manifests to chunks, wherein each forward-reference references a chunk by an identity of the chunk without identifying any server on which the chunk is stored; and

a chunk payload subsystem comprising a plurality of chunk server computers communicatively interconnected by the network, the chunk payload subsystem storing the chunks and encodes back-references to version manifests that reference the chunk, wherein each back-reference references an object by an identity of the object without identifying any server on which the object is stored,

further wherein the back-references maintained by the chunk payload subsystem comprise speculative back-references and verified back-references,

further wherein a speculative back-reference is created during a chunk put operation to put a chunk into the system, and the speculative back-reference is deleted if the speculative back-reference is of a lower class-of-storage than a verified back-reference from the chunk to the object.

2. The system of claim 1, wherein the speculative back-reference encodes a transaction identifier supplied with the chunk put operation.

3. The system of claim 1, wherein a put cancelling operation that identifies the chunk and the transaction identifier causes deletion of the speculative back-reference.

4. The system of claim 1, wherein the speculative back-reference further encodes a timestamp of the object put transaction.

5. The system of claim 1, wherein the speculative back-reference is created even when the chunk previously existed in the chunk payload subsystem.

6. The system of claim 1, wherein a verified back-reference corresponding to the speculative back-reference is created only after the manifest tracking subsystem has completed a transaction that included the chunk put operation.

7. The system of claim 1, wherein, in order to verify the speculative back-reference, the chunk payload subsystem queries the manifest tracking subsystem to confirm that there is a corresponding forward-reference to the chunk from the object.

8. The system of claim 1 further comprising:

a client which initiates an object put transaction,

wherein the client obtains a transaction identifier for the object put transaction from the manifest tracking subsystem by sending a request for transaction identifier that at least includes an identity of the object to be put.

9. The system of claim 8, wherein the request for transaction identifier includes a class-of-storage identifier.

10. The system of claim 8, wherein the manifest tracking subsystem cryptographically signs the transaction identifier, and wherein the transaction identifier encodes at least the identity of the object to be put and a unique transaction identifier.

11. The system of claim 10, wherein the transaction identifier is unique even when there are multiple concurrent transactions updating the object to be put.

12. The system of claim 8, wherein the chunk payload subsystem rejects a request for a chunk put operation if the request does not supply a transaction identifier which was cryptographically signed by the manifest subsystem.

13. The system of claim 1, wherein the manifest tracking subsystem and the chunk payload subsystem collaborate to maintain a class-of-storage attribute for each chunk stored in the system.

14. The system of claim 13, wherein the manifest tracking subsystem issues a notification of a class-of-storage update when a class-of-storage required for a chunk by an object has been updated.

15. The system of claim 14, wherein the chunk payload subsystem applies the class-of-storage update to a verified back-reference.

16. The system of claim 14, wherein the chunk payload subsystem creates the verified back-reference, if necessary, by confirming a speculative back-reference between the chunk and the object.

17. The system of claim 1, wherein a chunk get operation is validated using a transaction identifier obtained from the manifest tracking subsystem and supplied to the chunk payload subsystem with the chunk get operation.

18. The system of claim 17, wherein the chunk payload subsystem rejects the chunk get operation if the chunk does not have a back-reference to an object identified in the transaction identifier.

19. A system for providing distributed storage of data objects, the system comprising:

a manifest tracking subsystem comprising a plurality of manifest server computers communicatively interconnected by a network, the manifest tracking subsystem encoding forward-references from version manifests to chunks, wherein each forward-reference references a chunk by an identity of the chunk without identifying any server on which the chunk is stored; and

a chunk payload subsystem comprising a plurality of chunk server computers communicatively interconnected by the network, the chunk payload subsystem storing the chunks and encoding back-references to version manifests that reference the chunk, wherein each back-reference references an object by an identity of the object without identifying any server on which the object is stored,

further wherein the back-references maintained by the chunk payload subsystem comprise speculative back-references and verified back-references,

further wherein a speculative back-reference is created during a chunk put operation to put a chunk into the system, and, in order to verify the speculative back-reference, the chunk payload subsystem queries the manifest tracking subsystem to confirm that there is a corresponding forward-reference to the chunk from the object,

further wherein upon confirmation of the corresponding forward-reference, a verified back-reference from the chunk to the object is created if none already exists, and a class-of-storage attribute for the verified back-reference is updated if the verified back-reference already exists,

further wherein the speculative back-reference is then deleted.

20. The system of claim 19, wherein the class-of-storage attribute is updated to be a greater of the class-of-storage attributes in the speculative and the verified back-references.

21. A system for providing distributed storage of data objects, the system comprising:

a manifest tracking subsystem comprising a plurality of manifest server computers interconnected by a network, the manifest tracking subsystem encoding forward-references from version manifests to chunks, wherein each forward-reference references a chunk by an identity of the chunk without identifying any server on which the chunk is stored; and

a chunk payload subsystem comprising a plurality of chunk server computers interconnected by the network, the chunk payload subsystem storing the chunks and encoding back-references to version manifests that reference the chunk, wherein each back-reference references an object by an identity of the object without identifying any server on which the object is stored,

further wherein the back-references maintained by the chunk payload subsystem comprise speculative back-references and verified back-references,

further wherein a speculative back-reference is created during a chunk put operation to put a chunk into the system, and, in order to verify the speculative back-reference, the chunk payload subsystem queries the manifest tracking subsystem to confirm that there is a corresponding forward-reference to the chunk from the object,

further wherein upon confirmation of the corresponding forward-reference, a reference count for the chunk for the specific class-of-storage is incremented,

further wherein the speculative back-reference is then deleted.

22. The system of claim 21, wherein reference counts are adjusted in response to notifications from the manifest tracking subsystem that a class-of-storage for the chunk from a specific object has changed.

23. A method for providing storage of data objects using a distributed storage system, the method comprising:

encoding forward-references from version manifests to chunks in a manifest tracking subsystem which stores the version manifests, wherein each forward-reference references a chunk by an identity of the chunk without identifying any server on which the chunk is stored; and

encoding back-references to version manifests that reference the chunk, wherein each back-reference references an object by an identity of the object without identifying any server on which the object is stored,

further wherein the back-references maintained by the chunk payload system comprise speculative back-references and verified back-references,

further wherein a speculative back-reference is created during a chunk put operation to put a chunk into the system, and the speculative back-reference is deleted if the speculative back-reference is of a lower class-of-storage than a verified back-reference from the chunk to the object.