Deferred re-MRU operations to reduce lock contention

Patent metadata

Field	Value
Publication number	US-9733991-B2
Application number	US-201213616980-A
Country	US
Kind code	B2
Filing date	Sep 14, 2012
Priority date	Sep 14, 2012
Publication date	Aug 15, 2017
Grant date	Aug 15, 2017

Abstract

Data operations, requiring a lock, are batched into a set of operations to be performed on a per-core basis. A global lock for the set of operations is periodically acquired, the set of operations is performed, and the global lock is freed so as to avoid excessive duty cycling of lock and unlock operations in the computing storage environment.

First claim

The invention claimed is:

1. A method for data management in a multi-core computing storage environment by a processor device, comprising:
batching data operations, requiring a lock, into a set of operations to be performed on a per-core basis;
establishing a Least Recently Used (LRU) data list;
batching identification information of a portion of LRU data pursuant to a data hit, to be placed at a Most Recently Used (MRU) end of the LRU data list, as to determine whether data associated with the LRU data list should be demoted within the storage environment;
performing a demotion scan using a demote-ready list, the demote-ready list indicating data to be demoted within the storage environment due to no active data users;
periodically acquiring a global lock for the set of operations, performing the set of operations, and freeing the global lock so as to avoid excessive duty cycling of lock and unlock operations in the computing storage environment;
pursuant to batching the data operations, configuring at least one per-core array with each of a predetermined number of control blocks, and a flag associated with each of the predetermined number of control blocks; and
pursuant to batching the data operations, determining, in a first foreground process, if a batched array of the at least one per-core array is full, wherein:
if the batched array is full, an active list lock is acquired and the predetermined number of control blocks on the batched array are arranged on an MRU basis, and
if the batched array is not full, another one of the predetermined number of control blocks is added to the batched array, and a flag corresponding to the another one of the predetermined number of control blocks is set to indicate the another one of the predetermined number of control blocks is on the batched array; and

in a background process, flushing the at least one per-core array, including the batched array; wherein the background process increments a 2 bit counter of the at least one per-core array once every predetermined time interval while a second foreground process zeros the 2 bit counter each time the at least one per-core array, including the batched array, is flushed.

2. The method of claim 1 , further including performing the acquiring the global lock for the set of batched data operations based on a predetermined criteria.

3. The method of claim 1 , further including, if another flag corresponding to yet another one of the predetermined number of control blocks is set, scanning the at least one per-core array and removing the another one of the predetermined number of control blocks from the at least one per-core array, including at least the batched array.

4. A system for data management in a multi-core computing storage environment, comprising:
a processor device, operable in the multi-core computing storage environment, wherein the processor device:
batches data operations, requiring a lock, into a set of operations to be performed on a per-core basis;
establishes a Least Recently Used (LRU) data list;
batches identification information of a portion of LRU data pursuant to a data hit, to be placed at a Most Recently Used (MRU) end of the LRU data list, as to determine whether data associated with the LRU data list should be demoted within the storage environment;
performs a demotion scan using a demote-ready list, the demote-ready list indicating data to be demoted within the storage environment due to no active data users;
periodically acquires a global lock for the set of operations, performs performing the set of operations, and freeing the global lock so as to avoid excessive duty cycling of lock and unlock operations in the computing storage environment;
pursuant to batching the data operations, configures at least one per-core array with each of a predetermined number of control blocks, and a flag associated with each of the predetermined number of control blocks; and
pursuant to batching the data operations, determines, in a first foreground process, if a batched array of the at least one per-core array is full, wherein:
if the batched array is full, an active list lock is acquired and the predetermined number of control blocks on the batched array are arranged on an MRU basis, and
if the batched array is not full, another one of the predetermined number of control blocks is added to the batched array, and a flag corresponding to the another one of the predetermined number of control blocks is set to indicate the another one of the predetermined number of control blocks is on the batched array; and

in a background process, flushing the at least one per-core array, including the batched array; wherein the background process increments a 2 bit counter of the at least one per-core array once every predetermined time interval while a second foreground process zeros the 2 bit counter each time the at least one per-core array, including the batched array, is flushed.

5. The system of claim 4 , wherein the processor device performs the acquiring the global lock for the set of batched data operations based on a predetermined criteria.

6. The system of claim 4 , wherein the processor device, if another flag corresponding to yet another one of the predetermined number of control blocks is set, scans the at least one per-core array and removes the another one of the predetermined number of control blocks from the at least one per-core array, including at least the batched array.

7. The system of claim 4 , wherein the control blocks include Cache Directory Control Blocks (CDCBs).

8. The system of claim 4 , wherein the data operations to be batched include read hit workload in a cache of the computing storage environment.

9. A computer program product for data management in a multi-core computing storage environment, the computer program product comprising a non-transitory computer-readable storage medium having computer-readable program code portions stored therein, the computer-readable program code portions comprising:
a first executable portion that batches data operations, requiring a lock, into a set of operations to be performed on a per-core basis;
establishes a Least Recently Used (LRU) data list;
batches identification information of a portion of LRU data pursuant to a data hit, to be placed at a Most Recently Used (MRU) end of the LRU data list, as to determine whether data associated with the LRU data list should be demoted within the storage environment;
performs a demotion scan using a demote-ready list, the demote-ready list indicating data to be demoted within the storage environment due to no active data users;
a second executable portion that periodically acquires a global lock for the set of operations, performs the set of operations, and frees the global lock so as to avoid excessive duty cycling of lock and unlock operations in the computing storage environment;
a third executable portion that configures, pursuant to batching the data operations, at least one per-core array with each of a predetermined number of control blocks, and a flag associated with each of the predetermined number of control blocks; and
a fourth executable portion that, pursuant to batching the data operations, determines, in a first foreground process, if a batched array of the at least one per-core array is full, wherein:
if the batched array is full, an active list lock is acquired and the predetermined number of control blocks on the batched array are arranged on an MRU basis, and
if the batched array is not full, another one of the predetermined number of control blocks is added to the batched array, and a flag corresponding to the another one of the predetermined number of control blocks is set to indicate the another one

Assignees

• Ash Kevin John
• Benhase Michael Thomas
• Gupta Lokesh Mohan
• Todd Kenneth Wayne
• Whitworth David Blair
• IBM

Inventors

• Ash Kevin John
• Benhase Michael Thomas
• Gupta Lokesh Mohan
• Todd Kenneth Wayne
• Whitworth David Blair

Classifications

• G06F12/123

  with age lists, e.g. queue, most recently used [MRU] list or least recently used [LRU] list · CPC title

• G06F3/0659

  Command handling arrangements, e.g. command buffers, queues, command scheduling · CPC title

• G06F9/526 Primary

  Mutual exclusion algorithms · CPC title

• G06F12/0866

  for peripheral storage systems, e.g. disk cache · CPC title