[
https://issues.apache.org/jira/browse/HDFS-15094?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17009726#comment-17009726
]
Ayush Saxena commented on HDFS-15094:
-------------------------------------
Updated patch, to store ugi.toString() and reuse rather than recalculating for
hash and equals methods.
Did profiling for this, getConnection() prior used to take 1497 was reduced to
927.
A different approach thought was to use ugi.hashcode() for hash and
ugi.equals() in the equals methods, that in non secured environment reduced the
time to ~750, but in case of secured environment, the Subject of ugi is changed
at the router to set the user name, So in case of secured environment, each
call would create a new connection, as equals method of ugi, compares Subject,
which wont be same. So didn't use this approach. And limited to just storing
the ugi.toString().
> RBF : Reuse ugi string in ConnectionPoolID
> ------------------------------------------
>
> Key: HDFS-15094
> URL: https://issues.apache.org/jira/browse/HDFS-15094
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Ayush Saxena
> Assignee: Ayush Saxena
> Priority: Major
> Attachments: HDFS-15094-01.patch
>
>
> The connectionPoolID Hash Code and equals contains ugi.toString(), These
> methods are used as part of getConnection() in ConnectionManager as part of
> every call.
> The ugi.toString() eats up considerable amount of time, the Hash Calculation
> itself is ~10 percent of the total time of the call. And even more for the
> equals method.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
