Activity-based block management of a clustered file system

• US 9,015,210 B2
• Filed: 02/18/2014
• Issued: 04/21/2015
• Est. Priority Date: 08/15/2011
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Adkins, Janet, Brandyberry, Matthew T., Palsule, Ninad
• Primary Examiner(s): GOFMAN, ALEX N
• Application Number:
  US14/182,378
• Publication Number:
  US 20140164498A1
• Time in Patent Office: 427 Days
• Field of Search: 707/823, 707/781, 709/220, 709/223
• US Class Current: 707/823
• CPC Class Codes: G06F 16/1824 implemented using Network-attached Storage [NAS] architecture distributed or networked storage systems G06F3/067; protocols for distributed storage of data in a network H04L67/1097
• CPC Class Codes: G06F 16/1827 Management specifically adapted to NAS management of storage area networks [SAN] G06F3/067
• CPC Class Codes: H04L 12/6418 Hybrid transport
• CPC Class Codes: H04L 67/01 Protocols
• 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

A technique for operating a client node in a clustered file system includes allocating a number of blocks during a first time window and tracking the number of blocks allocated during the first time window. The technique further includes transmitting a block allocation request to a server node of the clustered file system for a number of requested blocks in response to a number of free blocks in a client-side block map reaching a first threshold value. In this case, the number of the requested blocks is based on the number of blocks allocated by the client node during the first time window.

Claims

1. A method of operating a client node in a clustered file system, the method comprising:

allocating, by the client node, a number of blocks during a first time window;

tracking, by the client node, the number of blocks allocated by the client node during the first time window;

transmitting, from the client node, a block allocation request to a server node of the clustered file system for a number of requested blocks in response to a number of free blocks in a client-side block map of one or more client-side block maps falling below a first threshold value, wherein the number of the requested blocks is based on the number of blocks allocated by the client node during the first time window;

receiving, at the client node, a client-side block map revocation message associated with the client-side block map from the server node;

in response to receiving the client-side block map revocation message, transmitting, from the client node, one or more extent messages to the server node, wherein the one or more extent messages each provide a contiguous set of blocks specified by a start block and a length that the client node is returning to the server node; and

transmitting, from the client node, a client-side block map response message to the server node of the clustered file system following transmission of a last one of the one or more extent messages, wherein the client-side block map response message indicates all blocks maintained by the client-side block map have been returned to the server node.

2. The method of claim 1, wherein the first threshold value is based on the number of the requested blocks.

3. The method of claim 1, further comprising:

initiating a block return from the client-side block map to the server node in response to the number of the requested blocks converging to zero and the free blocks in the client-side block map being greater than a minimum value.

4. The method of claim 1, wherein the client-side block map revocation message provides an indication to the client node to discontinue local block map management of a plurality of blocks in the client-side block map until local block map management is again allowed by the server node.

5. The method of claim 4, wherein:

the one or more client-side block maps comprise one client-side block map for each storage tier of a plurality of storage tiers;

each client-side block map of the one or more client-side block maps is associated with only one specific storage tier of the plurality of storage tiers; and

the client-side block map revocation message only revokes blocks in a first storage tier of the plurality of storage tiers.

6. The method of claim 1, further comprising:

determining whether the number of free blocks in the client-side block map is above the first threshold value;

in response to determining the number of free blocks in the client-side block map is above the first threshold value, determining whether the number of free blocks in the client-side block map is above a second threshold value, wherein the second threshold value is greater than the first threshold value; and

in response to determining whether the number of free blocks in the client-side block map is above the second threshold value, initiating the block return, wherein the block return returns one or more blocks that are not needed due to low block allocation activity of the one or more blocks.

7. The method of claim 1, wherein:

in response to a number of free blocks in the clustered file system increasing above a second threshold value, receiving, from the server node, a client-side block map initiation message, wherein the client-side block map initiation message indicates to the client node that the client node can resume caching the client-side block map.

8. A method of operating a server node in a clustered file system, the method comprising:

tracking, at the server node, a number of free blocks in the clustered file system;

in response the number of the free blocks in the clustered file system falling below a first threshold value, transmitting, from the server node, a client-side block map revocation message to one or more client nodes of the clustered file system, wherein the client-side block map revocation message provides an indication to one or more client nodes to discontinue local block map management until local block map management is again allowed by the server node;

determining whether the number of free blocks in the client-side block map is greater than the first threshold value;

in response to determining the number of free blocks in the client-side block map is greater than the first threshold value, determining whether the number of free blocks in the client-side block map is greater than a second threshold value, wherein the second threshold value is greater than the first threshold value; and

in response to determining the number of the free blocks in the clustered file system is greater than a second threshold value and receipt of a block allocate message from the first one of the client nodes, transmitting a client-side block map initiation message to a first one of the client nodes of the clustered file system, wherein in response to the first one of the client nodes receiving the client-side block map initiation, the first one of the client nodes initiates maintenance of the client-side block map.

9. The method of claim 8, wherein the client-side block map revocation message is broadcast to all of the client nodes in the clustered file system.

10. The method of claim 8, wherein the client-side block map revocation message is only transmitted to the client nodes in the clustered file system that are currently maintaining a client-side block map.

11. The method of claim 8, wherein the client-side block map revocation message only applies to a client-side block map for a specific storage tier of a plurality of storage tiers.

12. The method of claim 11, wherein:

the client-side block map is one of a plurality of client-side block maps;

each client-side block map of the one or more client-side block maps is associated with only one specific storage tier of the plurality of storage tiers; and

the client-side block map revocation message only revokes blocks in the specific storage tier.

13. The method of claim 11, wherein the client-side block map initiation message indicates to the client node that the client node can resume caching the client-side block map.