Delegated reference count base file versioning

• US 8,032,498 B1
• Filed: 06/29/2009
• Issued: 10/04/2011
• Est. Priority Date: 06/29/2009
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: Davenport, William, Bono, Jean-Pierre, Pawar, Sitaram, Veeraswamy, Sairam, Armangau, Philippe
• Primary Examiner(s): ORTIZ DITREN, BELIX M
• Application Number:
  US12/494,115
• Time in Patent Office: 827 Days
• Field of Search: None
• US Class Current: 707/690
• CPC Class Codes: G06F 16/128 Details of file system snapshots on the file-level, e.g. snapshot creation, administration, deletion use of snapshots for error detection or correction G06F11/14, G06F11/16

Abstract

A snapshot copy facility maintains information indicating block ownership and sharing between successive versions by delegating block reference counts to parent-child relationships between the file system blocks, as indicated by block pointers in inodes and indirect blocks. When a child block becomes shared between a parent block of the production file and a parent block of a snapshot copy, the delegated reference count is split among the parent blocks. This method is compatible with a conventional data de-duplication facility, and avoids a need to update block reference counts in block metadata of child blocks of a shared indirect block upon splitting the shared indirect block when writing to a production file.

Claims

1. A method of operating a network file server having a network adapter coupling the network file server to a data network of client computers, data storage storing files, and a data processor coupled to the data storage and the network adapter for access of the client computers to the files stored in the data storage, the files in the data storage being organized as a hierarchy of file system blocks including inodes, indirect blocks, and data blocks, the files stored in the data storage including a production file having a first data block in the hierarchy of file system blocks, the first data block being a child of the production file inode in the hierarchy of file system blocks, said method comprising:

storing in the data storage a reference count for the first data block, and storing in the data storage a delegated reference count for the parent-child relationship of the production file inode and the first data block;

the data processor creating in the data storage a snapshot copy of the production file by allocating a snapshot copy inode and sharing the first data block between the production file inode and the snapshot copy inode so that the first data block becomes a child of the snapshot copy inode, and splitting the delegated reference count between the parent-child relationship of the production file inode and the first data block and the parent-child relationship of the snapshot copy inode and the first data block; and

the data processor responding to a request from one of the client computers for writing data to the production file at an offset encompassed by the first data block by comparing the delegated reference count for the parent-child relationship of the production file inode and the first data block to the reference count for the first data block, and upon finding that the comparison indicates that the first data block is a shared block, allocating a second data block, writing to the second data block and decrementing the reference count for the first data block by the delegated reference count for the parent-child relationship of the production file inode and the first data block and linking the second data block to the production file inode in place of the first data block so that the second data block becomes a child of the production file inode and the first data block is no longer a child of the production file inode.

2. The method as claimed in claim 1, wherein the comparison indicates that the first data block is a shared block when the delegated reference count for the parent-child relationship of the production file inode and the first data block is not equal to the reference count for the first data block.

3. The method as claimed in claim 1, wherein the reference count for the first data block and the delegated reference count for the parent-child relationship of the production file inode and the first data block are set equal to a full-weight value before any snapshot copies have been made of the production file when the first data block is not shared with any other file in data storage, and wherein the delegated reference count is split between the parent-child relationship of the production file inode and the first data block and the parent-child relationship of the snapshot copy inode and the first data block by setting a delegated reference count for the parent-child relationship of the snapshot copy inode to a partial-weight value and decrementing the delegated reference count for the parent-child relationship of the production file inode and the first data block by the partial-weight value, wherein the partial-weight value is less than the full-weight value.

4. The method as claimed in claim 3, which further includes decrementing the reference count for the data block by the delegated reference count for the parent-child relationship of the snapshot copy inode and the first data block when the snapshot copy of the production file is deleted.

5. The method as claimed in claim 3, which further includes de-duplicating data of another file by changing a block pointer in an inode of said another file to point to the second data block, and incrementing the reference count for the second data block by the full-weight value.

6. The method as claimed in claim 1, wherein a pointer to the first data block is stored in the block pointer field of the production file inode, and the method further includes storing the delegated reference count for the parent-child relationship of the production file inode and the first data block in the block pointer field of the production file inode.

7. The method as claimed in claim 6, which further includes de-duplicating data of another file by changing a block pointer in an inode of said another file to point to the second data block, and incrementing the reference count for the second data block by a value of a delegated reference count in a block pointer field containing the block pointer in the inode of said another file.

8. The method as claimed in claim 1, wherein the production file also includes a first indirect block that is a child of the production file inode in the file system block hierarchy, and the method further includes the data processor responding to another request from said one of the client computers for writing data to the production file at an offset encompassed by the first indirect block by comparing a delegated reference count for the parent-child relationship of the production file inode and the first indirect block to a reference count for the first indirect block, and upon finding that the comparison indicates that the first indirect block is a shared block, allocating a second indirect block, copying content of the first indirect block to the second indirect block and decrementing the reference count for the first indirect block by the delegated reference count for the parent-child relationship of the production file inode and the first indirect block and linking the second indirect block to the production file inode in place of the first indirect block so that the second indirect block becomes a child of the production file inode and the second indirect block is no longer a child of the production file inode, and writing data to the production file at the offset encompassed by the first indirect block.

9. A method of operating a network file server having a network adapter coupling the network file server to a data network of client computers;

data storage storing files, and a data processor coupled to the data storage and the network adapter for access of the client computers to the files stored in the data storage, the files in the data storage being organized as a hierarchy of file system blocks including inodes, indirect blocks, and data blocks, the files stored in the data storage including a production file having a first indirect block in the hierarchy of file system blocks, the first indirect block being a child of the production file inode in the hierarchy of file system blocks, said method comprising;

storing in the data storage a reference count for the first indirect block, and storing in the data storage a delegated reference count for the parent-child relationship of the production file inode and the first indirect block,

the data processor creating in the data storage a snapshot copy of the production file by allocating a snapshot copy inode and sharing the first indirect block between the production file inode and the snapshot copy inode so that the first indirect block becomes a child of the snapshot copy inode, and splitting the delegated reference count between the parent-child relationship of the production file inode and the first indirect block and the parent-child relationship of the snapshot copy inode and the first indirect block by setting a delegated reference count for the parent-child relationship of the snapshot copy inode and the first indirect block to a value less than the delegated reference count for the parent-child relationship of the production file mode and the first indirect block and decrementing the delegated reference count for the parent-child relationship of the production file inode and the first indirect block by said value;

the data processor responding to a request from one of the client computers for writing data to the production file at an offset encompassed by the first indirect block by comparing the delegated reference count for the parent-child relationship of the production file inode and the first indirect block to the reference count for the first indirect block, and upon finding that the comparison indicates that the first indirect block is a shared block, allocating a second indirect block and copying content of the first indirect block to the second indirect block and decrementing the reference count for the first indirect block by the delegated reference count for the parent-child relationship of the production file inode and the first indirect block and linking the second indirect block to the production file inode in place of the first indirect block so that the second indirect block becomes a child of the production file inode and the first indirect block is no longer a child of the production file inode, and writing the data to the production file at the offset encompassed by the first indirect block.

10. The method as claimed in claim 9, wherein the comparison indicates that the first indirect block is a shared block when the delegated reference count for the parent-child relationship of the production file inode and the first indirect block is not equal to the reference count for the first indirect block.

11. The method as claimed in claim 9, wherein a pointer to the first indirect block is stored in a block pointer field of the production file inode, and wherein the method further includes storing the delegated reference count for the parent-child relationship of the production file inode and the first indirect block in the block pointer field of the production file inode.

12. A network file server comprising:

a network adapter for coupling the network file server to a data network of client computers;

data storage storing files; and

a data processor coupled to the data storage and the network adapter for access of the client computers to the files stored in the data storage;

the files in the data storage being organized as a hierarchy of file system blocks including inodes, indirect blocks, and data blocks, the files stored in the data storage including a production file having a first data block in the hierarchy of file system blocks, the first data block being a child of the production file inode in the hierarchy of file system blocks, and the data storage storing a reference count for the first data block, and the data storage storing a delegated reference count for the parent-child relationship of the production file inode and the first data block;

the data processor being programmed with a snapshot copy facility, the snapshot copy facility being executable by the data processor to create in the data storage a snapshot copy of the production file by allocating a snapshot copy inode and sharing the first data block between the production file inode and the snapshot copy inode so that the first data block becomes a child of the snapshot copy inode, and splitting the delegated reference count between the parent-child relationship of the production file inode and the first data block and the parent-child relationship of the snapshot copy inode and the first data block; and

the snapshot copy facility being executable by the data processor to respond to a request from one of the client computers for writing data to the production file at an offset encompassed by the first data block by comparing the delegated reference count for the parent-child relationship of the production file inode and the first data block to the reference count for the first data block, and upon finding that the comparison indicates that the first data block is a shared block, allocating a second data block, writing to the second data block and decrementing the reference count for the first data block by the delegated reference count for the parent-child relationship of the production file inode and the first data block and linking the second data block to the production file inode in place of the first data block so that the second data block becomes a child of the production file inode and the first data block is no longer a child of the production file inode.

13. The network file server as claimed in claim 12, wherein the comparison indicates that the first data block is a shared block when the delegated reference count for the parent-child relationship of the production file inode and the first data block is not equal to the reference count for the first data block.

14. The network file server as claimed in claim 12, wherein the reference count for the first data block and the delegated reference count for the parent-child relationship of the production file inode and the first data block are set equal to a full-weight value before any snapshot copies have been made of the production file when the first data block is not shared with any other file in data storage, and wherein the delegated reference count is split between the parent-child relationship of the production file inode and the first data block and the parent-child relationship of the snapshot copy inode and the first data block by setting a delegated reference count for the parent-child relationship of the snapshot copy inode to a partial-weight value and decrementing the delegated reference count for the parent-child relationship of the production file inode and the first data block by the partial-weight value, wherein the partial-weight value is less than the full-weight value.

15. The network file server as claimed in claim 14, wherein the snapshot copy facility is further executable by the data processor to decrement the reference count for the first data block by the delegated reference count for the parent-child relationship of the snapshot copy inode and the first data block when the snapshot copy of the production file is deleted.

16. The network file server as claimed in claim 14, wherein the data processor is further programmed with a data de-duplication facility, and the data de-duplication facility is executable by the data processor to de-duplicate data of another file by changing a block pointer in an inode of said another file to point to the second data block, and incrementing the reference count for the second data block by the full-weight value.

17. The network file server as claimed in claim 12, wherein a pointer to the first data block is stored in a block pointer field of the production file inode, and the delegated reference count for the parent-child relationship of the production file inode and the first data block also is stored in the block pointer field of the production file inode.

18. The network file server as claimed in claim 17, wherein the data processor is further programmed with a data de-duplication facility, and the data de-duplication facility is executable by the data processor to de-duplicate data of the production file by changing a block pointer in an indirect block of the production file to point to the first data block, and incrementing the reference count for the first data block by a value of a delegated reference count in a block pointer field containing the block pointer in the indirect block of the production file.

19. The network file server as claimed in claim 17, wherein the data processor is further programmed with a data de-duplication facility, and the data de-duplication facility is executable by the data processor to de-duplicate data of another file by changing a block pointer in an inode of said another file to point to the first data block, and incrementing the reference count for the first data block by a value of a delegated reference count in a block pointer field containing the block pointer in the inode of said another file.

20. The network file server as claimed in claim 12, wherein the production file also includes a first indirect block that is a child of the production file inode in the file system block hierarchy, and

the snapshot copy facility is further executable by the data processor to respond to another request from said one of the client computers for writing data to the production file at an offset encompassed by the first indirect block by comparing a delegated reference count for the parent-child relationship of the production file inode and the first indirect block to a reference count for the first indirect block, and upon finding that the comparison indicates that the first indirect block is a shared block, allocating a second indirect block, copying content of the first indirect block to the second indirect block and decrementing the reference count for the first indirect block by the delegated reference count for the parent-child relationship of the production file inode and the first indirect block and linking the second indirect block to the production file inode in place of the first indirect block so that the second indirect block becomes a child of the production file inode and the second indirect block is no longer a child of the production file inode, and writing data to the production file at the offset encompassed by the first indirect block.