[
https://issues.apache.org/jira/browse/HDDS-6402?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tanvi Penumudy updated HDDS-6402:
---------------------------------
Description:
The goal of this task is to allow more concurrent operations(mixture of write
and read) by replacing the global lock with a {{hash locking mechanism}} on key
path.
The performance of Ozone’s metadata write operations is limited by a {{{}bucket
lock{}}}, which is a *global lock* for key operations. Each metadata write
operation acquires the global bucket lock and holds it until that metadata
operation is finished. This prevents concurrent execution of write and read
operations even if they are fully independent on disjoint key paths.
In this task, we will be considering only the keys stored under
{{OBJECT_STORE("OBS")}} bucket layout format. In future, the plan is to refine
bucket locks for FSO bucket layouts as well, where we will need more complex
hierarchical structure prefix based locking mechanism.
was:
The goal of this task is to allow more concurrent operations(mixture of write
and read) by replacing the global lock with a {{hash locking mechanism}} on key
path.
The performance of Ozone’s metadata write operations is limited by a {{{}bucket
lock{}}}, which is a *global lock* for key operations. Each metadata write
operation acquires the global bucket lock and holds it until that metadata
operation is finished. This prevents concurrent execution of write and read
operations even if they are fully independent on disjoint key paths.
In this task, we will be considering only the keys stored under
{{OBJECT_STORE("OBS")}} bucket layout format. In future, the plan is to refine
bucket locks for FSO bucket layouts as well, where we will need more complex
hierarchical structure prefix based locking mechanism.
+Note:+ We will post the design doc soon.
> Introduce hashed key path locks for OBS bucket
> ----------------------------------------------
>
> Key: HDDS-6402
> URL: https://issues.apache.org/jira/browse/HDDS-6402
> Project: Apache Ozone
> Issue Type: New Feature
> Components: Ozone Manager
> Reporter: Rakesh Radhakrishnan
> Assignee: Tanvi Penumudy
> Priority: Major
> Attachments: KEY_PATH_LOCK Design Document.pdf
>
>
> The goal of this task is to allow more concurrent operations(mixture of write
> and read) by replacing the global lock with a {{hash locking mechanism}} on
> key path.
> The performance of Ozone’s metadata write operations is limited by a
> {{{}bucket lock{}}}, which is a *global lock* for key operations. Each
> metadata write operation acquires the global bucket lock and holds it until
> that metadata operation is finished. This prevents concurrent execution of
> write and read operations even if they are fully independent on disjoint key
> paths.
> In this task, we will be considering only the keys stored under
> {{OBJECT_STORE("OBS")}} bucket layout format. In future, the plan is to
> refine bucket locks for FSO bucket layouts as well, where we will need more
> complex hierarchical structure prefix based locking mechanism.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
