Github user aljoscha commented on the pull request:
https://github.com/apache/incubator-flink/pull/236#issuecomment-65228151
I ran some tests to measure the performance impact. I tried our regular
WordCount example, which uses Tuples, a POJO WordCount and a POJO WordCount
where the user function emits a subclass of the type of the DataSet.
The results:
- WordCount Tuple: 440 sec
- WordCount POJO (w/o this PR): 540 sec
- WordCount POJO (w/ this PR): 550 sec
- WordCount POJO Subclass: 2443 sec
- WordClass POJO Subclass with Tagging: 830 sec
The "tagging" version is a mockup of how it would behave if we allowed
users to register classes, so that only an ID has to be transferred. Similar to
how Kryo does it.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
