You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Zhipeng Zhang (Jira)" <ji...@apache.org> on 2023/03/13 04:18:00 UTC
[jira] [Updated] (FLINK-31410) ListStateWithCache Should support incremental snapshot
[ https://issues.apache.org/jira/browse/FLINK-31410?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zhipeng Zhang updated FLINK-31410:
----------------------------------
Description:
In Flink ML, we use ListStateWithCache [2] to enable caching data in memory and filesystem. However, it does not support incremental snapshot now — It writes all the data to checkpoint stream whenever calling snapshot [1], which could be inefficient.
[1][https://github.com/apache/flink-ml/blob/master/flink-ml-iteration/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/DataCacheSnapshot.java#L116]
[2][https://github.com/apache/flink-ml/blob/master/flink-ml-iteration/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/ListStateWithCache.java]
was:
In Flink ML, we used ListStateWithCache [2] to enable caching data in memory and filesystem. However, it does not support incremental snapshot now — It writes all the data to checkpoint stream when calling snapshot [1], which could be inefficient.
[1][https://github.com/apache/flink-ml/blob/master/flink-ml-iteration/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/DataCacheSnapshot.java#L116]
[2][https://github.com/apache/flink-ml/blob/master/flink-ml-iteration/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/ListStateWithCache.java]
> ListStateWithCache Should support incremental snapshot
> ------------------------------------------------------
>
> Key: FLINK-31410
> URL: https://issues.apache.org/jira/browse/FLINK-31410
> Project: Flink
> Issue Type: Improvement
> Components: Library / Machine Learning
> Affects Versions: ml-2.2.0
> Reporter: Zhipeng Zhang
> Priority: Major
>
> In Flink ML, we use ListStateWithCache [2] to enable caching data in memory and filesystem. However, it does not support incremental snapshot now — It writes all the data to checkpoint stream whenever calling snapshot [1], which could be inefficient.
>
> [1][https://github.com/apache/flink-ml/blob/master/flink-ml-iteration/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/DataCacheSnapshot.java#L116]
> [2][https://github.com/apache/flink-ml/blob/master/flink-ml-iteration/src/main/java/org/apache/flink/iteration/datacache/nonkeyed/ListStateWithCache.java]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)