[
https://issues.apache.org/jira/browse/SPARK-2011?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ankur Dave updated SPARK-2011:
------------------------------
Description: In the iteration loop, Pregel currently performs an innerJoin
to apply messages to vertices followed by an outerJoinVertices to join the
resulting subset of vertices back to the graph. These two operations could be
merged into a single call to joinVertices, which should be reimplemented in a
more efficient manner. This would allow us to examine only the vertices that
received messages. (was: In the iteration loop, Pregel currently performs an
innerJoin to apply messages to vertices followed by an outerJoinVertices to
join the resulting subset of vertices back to the graph. These two operations
could be merged into a single specialized operation called updateVertices. This
would allow us to examine only the vertices that received messages, and it
might eliminate a stage in every iteration.)
> Eliminate duplicate join in Pregel
> ----------------------------------
>
> Key: SPARK-2011
> URL: https://issues.apache.org/jira/browse/SPARK-2011
> Project: Spark
> Issue Type: Improvement
> Components: GraphX
> Reporter: Ankur Dave
> Assignee: Ankur Dave
>
> In the iteration loop, Pregel currently performs an innerJoin to apply
> messages to vertices followed by an outerJoinVertices to join the resulting
> subset of vertices back to the graph. These two operations could be merged
> into a single call to joinVertices, which should be reimplemented in a more
> efficient manner. This would allow us to examine only the vertices that
> received messages.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
