SCALABLE DISTRIBUTED STORAGE ARCHITECTURE

• US 20150058384A1
• Filed: 08/26/2013
• Published: 02/26/2015
• Est. Priority Date: 08/26/2013
• Status: Active Grant

• Current Assignee: Vmware LLC (Broadcom, Inc.)
• Original Assignee: VMware, Inc. (Broadcom, Inc.)
• Inventors: KARAMANOLIS, Christos, VASANI, Soam
• Granted Patent: US 9,811,531 B2
• Time in Patent Office: Days
• US Class Current: 707/827
• CPC Class Codes: G06F 16/182 Distributed file systems
• CPC Class Codes: G06F 16/188 Virtual file systems
• CPC Class Codes: G06F 16/27 Replication, distribution or synchronisation of data between databases or within a distributed database system; Distributed database system architectures therefor

Abstract

Techniques are disclosed for providing a file system interface for an object store intended to support simultaneous access to objects stored in the object store by multiple clients. In accordance with one method, an abstraction of a root directory to a hierarchical namespace for the object store is exposed to clients. The object store is backed by a plurality of physical storage devices housed in or directly attached to the plurality of host computers and internally tracks its stored objects using a flat namespace that maps unique identifiers to the stored objects. The creation of top-level objects appearing as subdirectories of the root directory is enabled, wherein each top-level object represents a separate abstraction of a storage device having a separate namespace that can be organized in accordance with any designated file system.

Claims

1. A method for providing a file system interface for an object store intended to support simultaneous access to objects stored in the object store by multiple clients, the steps comprising:

exposing to clients an abstraction of a root directory to a hierarchical namespace for the object store, wherein the object store (i) is backed by a plurality of physical storage devices housed in or directly attached to the plurality of host computers, and (ii) internally tracks its stored objects using a flat namespace that maps unique identifiers to the stored objects; and

enabling creation of top-level objects appearing as subdirectories of the root directory, wherein each top-level object represents a separate abstraction of a storage device having a separate namespace that can be organized in accordance with any designated file system.

2. The method of claim 1, wherein at least one top-level object is contains a clustered file system designed to support simultaneous access by clients up to a maximum limit.

3. The method of claim 2, wherein the top-level object is auto-mounted as a subdirectory at a client'"'"'s local version of the abstraction of the root directory when the client attempts to access a file stored within the clustered file system.

4. The method of claim 2, wherein the top-level object is configured to store metadata for file objects representing files stored in the clustered file system, wherein the file objects are stored in the object store separate from the top-level object.

5. The method of claim 4, wherein a file object is configured to store unique identifiers usable by the flat namespace to find locations of corresponding component objects that represent separate portions of the files and are stored in the object store separate from the file object.

6. The method of claim 4, wherein the metadata stored in the top-level object for a file object includes at least a file descriptor that comprises a unique identifier usable by the flat namespace to find a location of the file object in the object store.

7. The method of claim 6, wherein the file represented by the file object is a virtual disk file used by a virtual machine running on one of the plurality host computers.

8. The method of claim 7, wherein the top-level object is configured to store metadata for file objects that are only utilized by one of a plurality of virtual machines capable of being instantiated on any of the plurality of host computers.

9. The method of claim 3, wherein the top-level objects is automatically unmounted after a period of time in which no files stored within the clustered file system are accessed by the client.

10. The method of claim 1, wherein the flat namespace is implemented as an in-memory database, a synchronized copy of which is stored in the memory of each of the host computers.

11. The method of claim 1, wherein a pathname in the designated file system of a top level object can be mapped to an object configured to represent a separate abstraction of a storage device having a separate namespace that can be organized in accordance with any designated file system and can be auto-mounted at a client when the client attempts to access a file stored within the designated file system.

12. The method of claim 1, wherein an administrator specified name for an object is transformed into a unique identifier for the object when the flat namespace reports a conflict with a auto-generated unique identifier for the object.

13. A non-transitory computer readable storage medium storing instructions, which, when executed on a processor, performs an operation for providing a file system interface for an object store intended to support simultaneous access to objects stored in the object store by multiple clients, the operation comprising:

exposing an abstraction of a storage device to a plurality of host computers in a cluster, wherein the storage device is backed by an object store comprising, in aggregate, a plurality of physical storage devices housed in the plurality of host computers; and

enabling creation of multiple top-level file system objects in the object store that each represent separate file systems, wherein each top-level file system object is configured to (i) respond according to the semantics of its corresponding file system, and (ii) store metadata for file objects representing files stored in the corresponding file system, wherein the file objects are stored in the object store separate from their corresponding top-level file system objects.

14. The computer readable storage medium of claim 13, wherein at least one top-level object contains a clustered file system designed to support simultaneous access by clients up to a maximum limit.

15. The computer readable storage medium of claim 14, wherein the top-level object is auto-mounted as a subdirectory at a client'"'"'s local version of the abstraction of the root directory when the client attempts to access a file stored within the clustered file system.

16. The computer readable storage medium of claim 14, wherein the top-level object is configured to store metadata for file objects representing files stored in the clustered file system, wherein the file objects are stored in the object store separate from the top-level object.

17. The computer readable storage medium of claim 16, wherein a file object is configured to store unique identifiers usable by the flat namespace to find locations of corresponding component objects that represent separate portions of the files and are stored in the object store separate from the file object.

18. The computer readable storage medium of claim 16, wherein the metadata stored in the top-level object for a file object includes at least a file descriptor that comprises a unique identifier usable by the flat namespace to find a location of the file object in the object store.

21. The computer readable storage medium of claim 13, wherein the flat namespace is implemented as an in-memory database, a synchronized copy of which is stored in the memory of each of the host computers.

22. The computer readable storage medium of claim 13, wherein a pathname in the designated file system of a top level object can be mapped to an object configured to represent a separate abstraction of a storage device having a separate namespace that can be organized in accordance with any designated file system and can be auto-mounted at a client when the client attempts to access a file stored within the designated file system.

19. The computer readable storage medium of claim 19, wherein the file represented by the file object is a virtual disk file used by a virtual machine running on one of the plurality host computers.

20. The computer readable storage medium of claim 19, wherein the top-level object is configured to store metadata for file objects that are only utilized by one of a plurality of virtual machines capable of being instantiated on any of the plurality of host computers.

23. A computer system, comprising:

a processor and

a memory hosting a module, which, when executed on the processor, performs an operation for providing a file system interface for an object store intended to support simultaneous access to objects stored in the object store by multiple clients, the operation comprising;

exposing an abstraction of a storage device to a plurality of host computers in a cluster, wherein the storage device is backed by an object store comprising, in aggregate, a plurality of physical storage devices housed in the plurality of host computers; and

enabling creation of multiple top-level file system objects in the object store that each represent separate file systems, wherein each top-level file system object is configured to (i) respond according to the semantics of its corresponding file system, and (ii) store metadata for file objects representing files stored in the corresponding file system, wherein the file objects are stored in the object store separate from their corresponding top-level file system objects.

24. The computer system of claim 23, wherein at least one top-level object contains a clustered file system designed to support simultaneous access by clients up to a maximum limit.

25. The computer system of claim 24, wherein the top-level object is auto-mounted as a subdirectory at a client'"'"'s local version of the abstraction of the root directory when the client attempts to access a file stored within the clustered file system.

26. The computer system of claim 24, wherein the top-level object is configured to store metadata for file objects representing files stored in the clustered file system, wherein the file objects are stored in the object store separate from the top-level object.

27. The computer system of claim 26, wherein a file object is configured to store unique identifiers usable by the flat namespace to find locations of corresponding component objects that represent separate portions of the files and are stored in the object store separate from the file object.

28. The computer system of claim 26, wherein the metadata stored in the top-level object for a file object includes at least a file descriptor that comprises a unique identifier usable by the flat namespace to find a location of the file object in the object store.

29. The computer system of claim 28, wherein the file represented by the file object is a virtual disk file used by a virtual machine running on one of the plurality host computers.

30. The computer system of claim 29, wherein the top-level object is configured to store metadata for file objects that are only utilized by one of a plurality of virtual machines capable of being instantiated on any of the plurality of host computers.

31. The computer system of claim 23, wherein the flat namespace is implemented as an in-memory database, a synchronized copy of which is stored in the memory of each of the host computers.

32. The computer system of claim 23, wherein a pathname in the designated file system of a top level object can be mapped to an object configured to represent a separate abstraction of a storage device having a separate namespace that can be organized in accordance with any designated file system and can be auto-mounted at a client when the client attempts to access a file stored within the designated file system.