DISTRIBUTED CHANGE NOTIFICATIONS FOR A DISTRIBUTED FILESYSTEM

• US 20150356116A1
• Filed: 06/06/2014
• Published: 12/10/2015
• Est. Priority Date: 06/06/2014
• Status: Active Grant

• Current Assignee: Panzura, Inc.
• Original Assignee: Panzura, Inc.
• Inventors: Lin, Yun, Taylor, John Richard
• Granted Patent: US 9,646,022 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/184 implemented as replicated file system

Abstract

The disclosed embodiments disclose techniques for providing distributed change notifications for a distributed filesystem. Two or more cloud controllers collectively manage distributed filesystem data that is stored in the cloud storage systems; the cloud controllers ensure data consistency for the stored data, and each cloud controller caches portions of the distributed filesystem. During operation, a cloud controller receives a request from a client to access a file in the distributed filesystem. The cloud controller determines that the file needs strong read-after write consistency and sends a registration notification to the other cloud controllers of the distributed filesystem to request immediate notification of any changes to the file.

Claims

1. A computer-implemented method for requesting distributed change notifications for a distributed filesystem, the method comprising:

collectively managing the data of the distributed filesystem using two or more cloud controllers, wherein collectively managing the data comprises storing the data for the distributed filesystem in one or more cloud storage systems, wherein the cloud controllers cache and ensure data consistency for data stored in the cloud storage systems;

receiving at a first cloud controller a request from a client to access a file in the distributed filesystem;

determining that the file needs strong read-after-write consistency; and

sending a registration notification to one or more cloud controllers of the distributed filesystem, wherein the registration notification requests that the first cloud controller receive immediate notification of any changes to the file.

2. The computer-implemented method of claim 1,

wherein cloud controllers distribute updates for modified files that do not need strong read-after-write consistency via lazy incremental metadata snapshots;

wherein files that are being collaborated upon by multiple clients and/or client processes need strong read-after-write consistency; and

wherein changes to files that need strong read-after-write consistency need to be propagated to any clients accessing such files immediately after such files are written and closed.

3. The computer-implemented method of claim 2,

wherein the first cloud controller sends the registration notification to all of the other cloud controllers of the distributed filesystem;

wherein each cloud controller tracks the set of registration notification requests received from other cloud controllers for files in the distributed filesystem;

wherein a second cloud controller subsequently receives a file write and a subsequent file close request for the file from a second client; and

wherein the second cloud controller determines from its tracked set of registration notification requests that the first cloud controller has registered to receive change notifications for the modified file and sends a change notification message for the modified file to the first cloud controller.

4. The computer-implemented method of claim 3, wherein a given registration notification requests change notifications for a range of filesystem scopes that include:

a specific file;

a directory; and

a portion of the distributed filesystem hierarchy.

5. The computer-implemented method of claim 4,

wherein cloud controllers are configured to send change notification messages for a given modified file to only the set of cloud controllers that have previously sent notification requests with filesystem scopes that apply to the given modified file, thereby supporting strong read-after-write consistency while reducing the number of change notification messages that need to be sent between cloud controllers;

and

wherein cloud controllers that did not send notification requests for the given modified file are informed of updates to the given modified file lazily via a subsequent incremental metadata snapshot.

6. The computer-implemented method of claim 4, wherein sending the given registration notification with a coarse granularity of filesystem scope facilitates supporting strong read-after-write consistency for multiple files while reducing the number of registration notifications that are broadcast to all of the cloud controllers of the distributed filesystem.

7. The computer-implemented method of claim 3, wherein determining that the file needs strong read-after-write consistency involves one or more of the following:

determining from the file'"'"'s extension, file type, and/or location in the filesystem hierarchy that the file needs strong read-after-write consistency;

using a set of heuristic-based techniques that analyze file contents to determine a file type and/or predict access behavior to determine that the file needs strong read-after-write consistency; and

determining from a locality policy that the file needs strong read-after-write consistency.

8. The computer-implemented method of claim 3, wherein determining that the file needs strong read-after-write consistency involves tracking client file accesses to dynamically identify files that are being collaboratively accessed by one or more clients.

9. The computer-implemented method of claim 3,

wherein registration notifications are associated with an expiration time interval;

and

wherein cloud controllers track the expiration time interval for each received registration notification and retire expired registration notifications.

10. The computer-implemented method of claim 9, wherein the method further involves:

tracking the expiration time interval for the registration notification at the first cloud controller;

determining that the client will access the file beyond the expiration time interval; and

sending a subsequent registration notification for the file from the first cloud controller to all of the other cloud controllers for the distributed filesystem prior to the expiration of the first registration notification to ensure that no change notifications are missed between the expiration of the first registration notification and the receipt of the subsequent registration notification.

11. The computer-implemented method of claim 3,

wherein the second cloud controller receives a request to perform a namespace operation for the file from the client;

and

wherein the second cloud controller is configured to send a subsequent change notification message for the namespace operation to all of the cloud controllers that have registered to receive change notifications for the file.

12. The computer-implemented method of claim 3, wherein the method further comprises sending a synchronization update request from the first cloud controller to the second cloud controller to retrieve the modifications made to the file by the second client.

13. A non-transitory computer-readable storage medium storing instructions that when executed by a computer cause the computer to perform a method for requesting distributed change notifications for a distributed filesystem, the method comprising:

collectively managing the data of the distributed filesystem using two or more cloud controllers, wherein collectively managing the data comprises storing the data for the distributed filesystem in one or more cloud storage systems, wherein the cloud controllers cache and ensure data consistency for data stored in the cloud storage systems;

receiving at a first cloud controller a request from a client to access a file in the distributed filesystem;

determining that the file needs strong read-after-write consistency; and

sending a registration notification to one or more cloud controllers of the distributed filesystem, wherein the registration notification requests that the first cloud controller receive immediate notification of any changes to the file.

14. The non-transitory computer-readable storage medium of claim 13,

wherein cloud controllers distribute updates for modified files that do not need strong read-after-write consistency via lazy incremental metadata snapshots;

wherein files that are being collaborated upon by multiple clients and/or client processes need strong read-after-write consistency; and

wherein changes to files that need strong read-after-write consistency need to be propagated to any clients accessing such files immediately after such files are written and closed.

15. The non-transitory computer-readable storage medium of claim 14,

wherein the first cloud controller sends the registration notification to all of the other cloud controllers of the distributed filesystem;

wherein each cloud controller tracks the set of registration notification requests received from other cloud controllers for files in the distributed filesystem;

wherein a second cloud controller subsequently receives a file write and a subsequent file close request for the file from a second client; and

wherein the second cloud controller determines from its tracked set of registration notification requests that the first cloud controller has registered to receive change notifications for the modified file and sends a change notification message for the modified file to the first cloud controller.

16. The non-transitory computer-readable storage medium of claim 15, wherein a given registration notification requests change notifications for a range of filesystem scopes that include:

a specific file;

a directory; and

a portion of the distributed filesystem hierarchy.

17. The non-transitory computer-readable storage medium of claim 16,

wherein cloud controllers are configured to send change notification messages for a given modified file to only the set of cloud controllers that have previously sent notification requests with filesystem scopes that apply to the given modified file, thereby supporting strong read-after-write consistency while reducing the number of change notification messages that need to be sent between cloud controllers;

and

wherein cloud controllers that did not send notification requests for the given modified file are informed of updates to the given modified file lazily via a subsequent incremental metadata snapshot.

18. The non-transitory computer-readable storage medium of claim 15,

wherein registration notifications are associated with an expiration time interval;

and

wherein cloud controllers track the expiration time interval for each received registration notification and retire expired registration notifications.

19. The non-transitory computer-readable storage medium of claim 18, wherein the method further involves:

tracking the expiration time interval for the registration notification at the first cloud controller;

determining that the client will access the file beyond the expiration time interval; and

sending a subsequent registration notification for the file from the first cloud controller to all of the other cloud controllers for the distributed filesystem prior to the expiration of the first registration notification to ensure that no change notifications are missed between the expiration of the first registration notification and the receipt of the subsequent registration notification.

20. A cloud controller that requests distributed change notifications for a distributed filesystem, comprising:

a processor;

a storage mechanism that stores metadata for the distributed filesystem; and

a storage management mechanism;

wherein two or more cloud controllers collectively manage the data of the distributed filesystem;

wherein the cloud controller is configured to receive a request from a client to access a file in the distributed filesystem;

wherein the storage management mechanism determines that the file needs strong read-after-write consistency; and

wherein the storage management mechanism is configured to send a registration notification to one or more cloud controllers of the distributed filesystem to request that the first cloud controller receive immediate notification of any changes to the file.