Method for efficient delivery of clustered data via adaptive TCP connection migration

• US 7,565,446 B2
• Filed: 08/27/2007
• Issued: 07/21/2009
• Est. Priority Date: 08/27/2007
• Status: Active Grant

• Current Assignee: Innovations In Memory LLC
• Original Assignee: Gear Six, Inc.
• Inventors: Talagala, Nisha, Patterson, Martin
• Primary Examiner(s): Shin, Christopher B
• Application Number:
  US11/845,679
• Publication Number:
  US 20090059862A1
• Time in Patent Office: 694 Days
• Field of Search: 709238-245, 709/207, 370328-331, 710 5- 6
• US Class Current: 709/238
• CPC Class Codes: H04L 67/1097 for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]

Abstract

The present invention relates generally to a method for efficient I/O handling in a cluster-based architecture. According to one aspect, the invention enables efficient scheduling of TCP connection migrations within a cluster. According to another aspect, the invention enables I/Os performed as TCP handoff operations to coexist on the same TCP/IP connection with I/Os performed as remote operations.

Claims

1. A method comprising:

receiving a network connection at a first node of a storage cluster;

receiving a first I/O request at the first node via the connection;

processing the first I/O request at the first node;

receiving a second I/O request at the first node via the connection;

forwarding the second I/O request to a second node of the storage cluster;

processing the second I/O request at the second node while maintaining the network connection at the first node; and

migrating the connection to the second node after completing the processing of the first I/O request at the first node and after forwarding the second I/O request to the second node.

2. A method according to claim 1, further comprising:

logging statistics corresponding to the first and second I/O requests; and

determining when to migrate the connection to the second node based on an evaluation of the statistics.

3. A method according to claim 1, further comprising:

maintaining a directory of data associated with the storage cluster; and

identifying the second node using the directory.

4. A method according to claim 1, wherein the step of migrating the connection is performed transparently to a client associated with the connection.

5. A method according to claim 1, wherein the connection uses TCP.

6. A method according to claim 4, wherein the connection uses TCP.

7. A method according to claim 1, further comprising sending a reply to the first I/O request from the first node before migrating the connection.

8. A method according to claim 7, further comprising sending a reply to the second I/O request from the second node after migrating the connection.

9. A method comprising:

receiving a network connection at a first node of a storage cluster;

receiving a first I/O request at the first node via the connection;

processing the first I/O request at the first node;

receiving a second I/O request at the first node via the connection;

determining whether to handoff the second I/O request to a second node of the storage cluster;

if the determination is to handoff the second I/O request;

processing the second I/O request at the second node, and

migrating the connection to the second node after completing the processing of the first I/O request at the first node and after forwarding the second I/O request to the second node;

if the determination is to not handoff the second I/O request;

processing the second I/O request at the second node, and

forwarding results of processing the second I/O request to the first node.

10. A method according to claim 9, further comprising:

maintaining a directory of data associated with the storage cluster; and

identifying the second node using the directory.

11. A method according to claim 9, wherein the step of migrating the connection is performed transparently to a client associated with the connection.

12. A method according to claim 9, wherein the connection uses TCP.

13. A method according to claim 11, wherein the connection uses TCP.

14. A storage apparatus comprising:

a first node that receives a network connection and a first I/O request via the connection;

a second node that processes a second I/O request received at the first node while maintaining the network connection at the first node; and

a handoff mechanism that migrates the connection to the second node after processing of the first I/O request at the first node is completed and after processing of the second I/O request at the second node has started.

15. An apparatus according to claim 14, further comprising:

statistics corresponding to the processing of the first and second I/O requests,

wherein the handoff mechanism uses the statistics to determine when to migrate the connection to the second node.

16. An apparatus according to claim 14, further comprising:

a directory of data associated with the storage cluster,

wherein the second node is identified using the directory.

17. An apparatus according to claim 14, wherein the handoff mechanism is adapted to migrate the connection transparently to a client associated with the connection.

18. An apparatus according to claim 14, wherein the connection uses TCP.

19. An apparatus according to claim 17, wherein the connection uses TCP.

20. An apparatus according to claim 14, wherein the handoff mechanism is further adapted to determine whether to handoff the second I/O request to a second node of the storage cluster, and wherein if the determination is to handoff the second I/O request:

the handoff mechanism causes the second I/O request to be processed at the second node, and migrates the connection to the second node after processing of the first I/O request at the first node is completed and after processing of the second I/O request at the second node has started;

and wherein if the determination is to not handoff the second I/O request;

the handoff mechanism causes the second I/O request to be processed at the second node, and further causes results of processing the second I/O request to be forwarded to the first node.