[
https://issues.apache.org/jira/browse/SPARK-1172?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13954787#comment-13954787
]
Andrew Or commented on SPARK-1172:
----------------------------------
Here's a quick clarification of the current naming scheme:
BlockMaster runs on both executors the driver. On the driver, it's only used in
the local mode.
BlockManagerMaster runs on both executors and the driver. It handles the
communication from each BlockManager to the BlockManagerMasterActor. On the
driver, this is a local inter-thread communication.
BlockManagerMasterActor exists only on the driver. It maintains an overview of
which blocks are where and other useful information. It does not necessarily
know about all blocks, however, because there is an option in the BlockManager
to update a block without telling master (i.e. BlockManagerMasterActor).
BlockManagerSlaveActor exists on both executors and the driver. It handles the
communication from the BlockManagerMasterActor to each BlockManager. As before,
this is a local inter-thread communication on the driver.
> Improve naming of the BlockManager classes
> ------------------------------------------
>
> Key: SPARK-1172
> URL: https://issues.apache.org/jira/browse/SPARK-1172
> Project: Apache Spark
> Issue Type: Improvement
> Components: Block Manager
> Reporter: Patrick Wendell
> Assignee: Andrew Or
> Priority: Minor
> Fix For: 1.0.0
>
>
> There is something called BlockManagerMaster that doesn't run on the driver
> (only on the Executors). There is a thing called BlockManagerWorker that is
> not really a "worker" component but a bridge between the connection manager
> and the block manager. I found these names confusing on my first read of this
> code and in the last week two other people have asked me to explain it to
> them. So perhaps we can find more intuitive names here... :)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
