[
https://issues.apache.org/jira/browse/HDFS-14403?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16812852#comment-16812852
]
Christopher Gregorian commented on HDFS-14403:
----------------------------------------------
Thanks for the comments and suggestions! Patch 002 should fix up those tests
and the checkstyle warnings, as well as address [~xkrogen]'s suggestions: I
built upon those by switching to using a {{LongAdder}} in the internal
{{UserLockHoldTime}} class instead of an {{AtomicLong}}, and chose to use 5 as
the default write time multiplier.
I approached [~linyiqun]'s second suggestion by changing the
{{DecayRpcProvider}} to update from the {{CostProvider}} on each call only when
there isn't a value in the cache, and keeping the update on decay sweep: I
believe this is a good compromise in minimizing contention on the
{{FSNamesystemLock}}'s {{userLockTimes}} while still keeping the totals up to
date for new users. Let me know if there's a reason I'm not thinking of where
updating every call is preferable :)
> Cost-Based RPC FairCallQueue
> ----------------------------
>
> Key: HDFS-14403
> URL: https://issues.apache.org/jira/browse/HDFS-14403
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: ipc, namenode
> Reporter: Erik Krogen
> Assignee: Christopher Gregorian
> Priority: Major
> Labels: qos, rpc
> Attachments: CostBasedFairCallQueueDesign_v0.pdf, HDFS-14403.001.patch
>
>
> HADOOP-15016 initially described extensions to the Hadoop FairCallQueue
> encompassing both cost-based analysis of incoming RPCs, as well as support
> for reservations of RPC capacity for system/platform users. This JIRA intends
> to track the former, as HADOOP-15016 was repurposed to more specifically
> focus on the reservation portion of the work.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
