[
https://issues.apache.org/jira/browse/FLINK-11986?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Yu Li updated FLINK-11986:
--------------------------
Component/s: Benchmarks
> Add micro benchmark for state operations
> ----------------------------------------
>
> Key: FLINK-11986
> URL: https://issues.apache.org/jira/browse/FLINK-11986
> Project: Flink
> Issue Type: Improvement
> Components: Benchmarks, Runtime / State Backends, Tests
> Reporter: Yu Li
> Assignee: Yu Li
> Priority: Major
> Fix For: 1.9.0
>
>
> Currently in the
> [flink-benchmark|https://github.com/dataArtisans/flink-benchmarks] project we
> already have a JMH case for the whole backend, but none for finer grained
> state operations, and here we propose adding JMH cases for them, including
> (but not limited to):
> * ValueState
> ** testPut
> ** testGet
> * ListState
> ** testUpdate
> ** testGet
> ** testAddAll
> * MapState
> ** testPut
> ** testGet
> ** testContains
> ** testKeys
> ** testValues
> ** testEntries
> ** testIterator
> ** testRemove
> ** testPutAll
> And we will create benchmark for {{HeapKeyedStateBackend}} and
> {{RocksDBKeyedStateBackend}} separately.
> We believe these micro benchmarks could help locate where the regression
> comes from in case any observed from the backend benchmark, and we could also
> use these benchmarks to assure no performance downgrade when modifying
> relative codes.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
