ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-555809927
close it.
This is an automated message
ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-547238238
I have tested it locally, the performance is almost equally:
Test case:
```scala
object
ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-546806343
I tested simply:
`./spark-shell --executor-cores 36 --executor-memory 120G
--total-executor-cores
ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-543579149
Yeah, it can provide great performance improvement from end to end. The
single record
ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-529342191
I'm sorry for the late reply. The end-to-end perf improvement should be
small because of the RPC has a
ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-525296971
This found in the RPC module which uses java serializer. Not sure why can't
configure with `Kryo`. And
ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-525114090
Yeah, it has the state:
```java
/** wire handle -> obj/exception map */
private final HandleTable
ConeyLiu commented on issue #25470: [SPARK-28751][Core][WIP] Improve java
serializer deserialization performance
URL: https://github.com/apache/spark/pull/25470#issuecomment-524615974
Hi @cloud-fan, it's difficult to reuse the `ObjectInpuStream`. The workflow
of `ObjectOutputStream` write