Github user CodingCat commented on the pull request:
https://github.com/apache/spark/pull/214#issuecomment-38556226
Hey, @qqsun8819 , after the second thought on whether task serialization
function should call the function directly or send a message to the
ClusterSchedulerBackend, I think sending a message is a better choice in my
view, not only for keeping consistency with the current implementation
if you call the function directly, the thread will only return after the
LaunchTask message is sent to the Executor, potentially prolonging the queuing
time of other serialize tasks, if you use fixed threadpool (if you use
cachedthreadpool, it will slow down the resource release speed). Also, calling
the function in multiple external threads may bring more concurrency issues...
---
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.
---
