[
https://issues.apache.org/jira/browse/HDFS-17603?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17882955#comment-17882955
]
ASF GitHub Bot commented on HDFS-17603:
---------------------------------------
kokon191 opened a new pull request, #7054:
URL: https://github.com/apache/hadoop/pull/7054
A `LockPoolManager` that stores locks and a ref count. Ref count increments
upon lock acquisition, decrements upon lock release. Lock is ejected when ref
count hits 0. Cached locks get an extra ref count and won't be ejected by lock
releases. Cached locks are updated periodically with a `PromotionService`,
engined by the `PromotionMonitor` that tracks which locks are the most active.
> [FGL] Abstract a LockManager to manage locks
> ---------------------------------------------
>
> Key: HDFS-17603
> URL: https://issues.apache.org/jira/browse/HDFS-17603
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: ZanderXu
> Assignee: Felix N
> Priority: Major
> Attachments: image-2024-08-14-17-17-28-792.png
>
>
> Abstract a LockManager to manage locks.
> Some requirements for this LockManager:
> * Cached a fixed number lock instances, such as: 1000
> * Assign a Lock instance to a key and keep this mapping until the key
> released this instance
> * This LockManager needs a high performance, such as: QPS 1000w
>
> Some implementations that we can refer to:
> * alluxio.collections.LockPool in Alluxio
> * Implementation in MEITUAN
> !image-2024-08-14-17-17-28-792.png|width=205,height=196!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
