Github user mxm commented on the pull request:
https://github.com/apache/flink/pull/1125#issuecomment-140022781
In addition to the discussion in the JIRA, I had to change the way the
network memory is configured because otherwise I cannot calculate the direct
memory size correctly. The default network buffer size is not easy to get in
the startup scripts without also putting it in the config. Besides that, I
think it will be much easier for users to configure the network buffer size
using an absolute value rather than some obscure numberOfNetworkBuffers.
> We have to be clear and careful with the term managed memory also, as the
regular JVM heap is managed memory (managed by the JVM / GC) as opposed to the
unmanaged memory in a C++ program.
In Flink, we always use the term managed memory to refer to our own memory
management. So I think it makes sense to stick to that term. The former name
`taskmanager.memory.size` was not clearer in any way because one would think
that this refers to the total memory size rather than the Flink managed memory
size.
---
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.
---
