Intelligent selection of replication node for file data blocks in GPFS-SNC

• US 9,471,586 B2
• Filed: 01/10/2013
• Issued: 10/18/2016
• Est. Priority Date: 01/10/2013
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Patil, Sandeep R., Sarkar, Prasenjit, Shiraguppi, Riyazahamad M., Sivakumar, Gandhi
• Primary Examiner(s): Trujillo, James
• Assistant Examiner(s): Khoshnoodi, Fariborz
• Application Number:
  US13/738,033
• Publication Number:
  US 20140195488A1
• Time in Patent Office: 1,377 Days
• Field of Search: 707/634, 707/610
• US Class Current: 1/1
• CPC Class Codes: G06F 16/1752 based on file chunks

Abstract

A mechanism is provided in a data processing system for replicating writing of a file with striping. The mechanism writes a file at an owner node within a plurality of nodes in a data processing system. The mechanism divides the file into a plurality of file chunks. The mechanism identifies at least one replication node within the plurality of nodes having a duplicate copy of a respective file chunk within the plurality of file chunks. The mechanism selects a plurality of replication nodes for the plurality of file chunks based on identification at least one replication node within the plurality of nodes having a duplicate copy of a respective file chunk and replicates the file at the plurality of replication nodes based on the selection of the plurality of replication nodes.

Claims

1. A method, in a data processing system, for replicating writing of a file with striping, the method comprising:

writing a file at an owner node within a plurality of nodes in a data processing system;

dividing the file into a plurality of file chunks;

performing a hash function on each file chunk to form a hash value;

looking up each hash value in a hash repository;

identifying at least one replication node within the plurality of nodes having a duplicate copy of a respective file chunk within the plurality of file chunks responsive to a hash value of the respective file chunk having an entry in the hash repository;

adding an entry to a duplicate set for each file chunk for which a duplicate copy is already stored at a replication node and adding an entry to a no-duplicate set for each file chunk for which no duplicate copy is stored at a replication node;

selecting a plurality of replication nodes, including the at least one replication node, for the plurality of file chunks based on the duplicate set and the no-duplicate set; and

replicating the file at the plurality of replication nodes based on the selection of the plurality of replication nodes, wherein replicating the file at the plurality of replication nodes comprises transferring pointer information for each file chunk having an entry in the duplication list to a corresponding selected replication node, wherein the pointer information points to the copy of the file chunk at the corresponding selected replication node.

2. The method of claim 1, wherein writing a file at an owner node comprises:

receiving a request to write a file at an upper layer in the data processing system;

selecting, by the upper layer, the owner node from the plurality of nodes; and

writing, by the upper layer, the file to the owner node, wherein the owner node writes the file to one or more local disks that are local to the owner node.

3. The method of claim 2, wherein the local node does not share the one or more local disks with others of the plurality of nodes.

4. The method of claim 1, wherein performing the hash function on each file chunk is performed by a common de-duplication engine in an upper layer in the data processing system.

5. The method of claim 1, wherein performing the hash function on each file chunk is performed by a local hash engine at the owner node.

6. The method of claim 1, wherein the hash repository is within an upper layer in the data processing system.

7. The method of claim 1, wherein identifying at least one replication node having a duplicate copy of a respective file chunk comprises passing the duplicate set and the no-duplicate set to the owner node.

8. The method of claim 1, wherein selecting a plurality of replication nodes for the plurality of file chunks comprises selecting a replication node within the plurality of nodes for each file chunk having an entry in the no-duplicate set based on replication data balancing requirements such that file data are uniform across the plurality of nodes.

9. The method of claim 1, wherein replicating the file at the plurality of replication nodes comprises creating, by the selected replication node, a copy of the file chunk data.

10. A computer program product comprising a non-transitory computer readable storage medium having a computer readable program stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to:

write a file at an owner node within a plurality of nodes in a data processing system;

divide the file into a plurality of file chunks;

perform a hash function on each file chunk to form a hash value;

look up each hash value in a hash repository;

identify at least one replication node within the plurality of nodes having a duplicate copy of a respective file chunk within the plurality of file chunks responsive to a hash value of the respective file chunk having an entry in the hash repository;

add an entry to a duplicate set for each file chunk for which a duplicate copy is already stored at a replication node and add an entry to a no-duplicate set for each file chunk for which no duplicate copy is stored at a replication node;

select a plurality of replication nodes, including the at least one replication node, for the plurality of file chunks based on the duplicate set and the no-duplicate set; and

replicate the file at the plurality of replication nodes based on the selection of the plurality of replication nodes, wherein replicating the file at the plurality of replication nodes comprises transferring pointer information for each file chunk having an entry in the duplication list to a corresponding selected replication node wherein the pointer information points to the copy of the file chunk at the corresponding selected replication node.

11. The computer program product of claim 10, wherein writing a file at an owner node comprises:

receiving a request to write a file at an upper layer in the data processing system;

selecting, by the upper layer, the owner node from the plurality of nodes; and

writing, by the upper layer, the file to the owner node, wherein the owner node writes the file to one or more local disks that are local to the owner node.

12. The computer program product of claim 10, wherein performing the hash function on each file chunk is performed by a common de-duplication engine in an upper layer in the data processing system.

13. The computer program product of claim 10, wherein performing the hash function on each file chunk is performed by a local hash engine at the owner node.

14. The computer program product of claim 11, wherein the local node does not share the one or more local disks with others of the plurality of nodes.

15. The computer program product of claim 10, wherein the hash repository is within an upper layer in the data processing system.

16. The computer program product of claim 10, wherein identifying at least one replication node having a duplicate copy of a respective file chunk comprises passing the duplicate set and the no-duplicate set to the owner node.

17. The computer program product of claim 10, wherein selecting a plurality of replication nodes for the plurality of file chunks comprises selecting a replication node within the plurality of nodes for each file chunk having an entry in the no-duplicate set based on replication data balancing requirements such that file data are uniform across the plurality of nodes.

18. The computer program product of claim 10, wherein replicating the file at the plurality of replication nodes comprises creating, by the selected replication node, a copy of the file chunk data.

19. An apparatus, comprising:

a processor; and

a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to;

write a file at an owner node within a plurality of nodes in a data processing system;

divide the file into a plurality of file chunks;

perform a hash function on each file chunk to form a hash value;

look up each hash value in a hash repository;

identify at least one replication node within the plurality of nodes having a duplicate copy of a respective file chunk within the plurality of file chunks responsive to a hash value of the given file chunk having an entry in the hash repository responsive to a hash value of the respective file chunk having an entry in hash repository;

add an entry to a duplicate set for each file chunk for which a duplicate copy is already stored at a replication node and add an entry to a no-duplicate set for each file chunk for which no duplicate copy is stored at a replication node;

select a plurality of replication nodes, including the at least one replication node, for the plurality of file chunks based on the duplicate set and the no-duplicate set; and

replicate the file at the plurality of replication nodes based on the selection of the plurality of replication nodes, wherein replicating the file at the plurality of replication nodes comprises transferring pointer information for each file chunk having an entry in the duplication list to a corresponding selected replication node, wherein the pointer information points to the copy of the file chunk at the corresponding selected replication node.

20. The apparatus of claim 19, wherein performing the hash function on each file chunk is performed using a common de-duplication engine in an upper layer in the data processing system.

21. The apparatus of claim 19, wherein performing the hash function on each file chunk is performed using a local hash engine at the owner node.

22. The apparatus of claim 19, wherein writing a file at an owner node comprises:

receiving a request to write a file at an upper layer in the data processing system;

selecting, by the upper layer, the owner node from the plurality of nodes; and

writing, by the upper layer, the file to the owner node, wherein the owner node writes the file to one or more local disks that are local to the owner node.

23. The apparatus of claim 22, wherein the local node does not share the one or more local disks with others of the plurality of nodes.

24. The apparatus of claim 19, wherein the hash repository is within an upper layer in the data processing system.

25. The apparatus of claim 19, wherein replicating the file at the plurality of replication nodes comprises creating, by the selected replication node, a copy of the file chunk data.