[
https://issues.apache.org/jira/browse/HDFS-14403?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16822129#comment-16822129
]
Chao Sun commented on HDFS-14403:
---------------------------------
Thanks [~cgregori]. I like the way how these two patches getting combined
(except that we should split them into two parts). A few comments on the patch
v5:
- In {{ProcessingDetails}}, when is {{ENQUEUE}} be set?
- In {{ProcessingDetails}}, should we use nano seconds as the default time
unit as oppose to micro seconds?
- In {{RpcMetrics}}, the {{@Metric}} for lock wait time should be "Lock wait
time", not "Processing time".
- In {{Server#updateMetrics}}, we don't really need to pass in {{timestamp}} -
it can be obtained from {{call}}.
- In {{DecayRpcScheduler#addResponseTime}} and a few other places we are
casting long to int and then back to long later. It's better if we can just use
long all the way.
- We should update the doc for {{DecayRpcScheduler#getAndIncrementCost}} to
reflect the new changes.
- We need class level comments for {{ProcessingDetails}} to explain what it is
for.
- nit: better to align the comments on {{ProcessingDetails.Timing}}.
- nit: we don't need a {{VALUES}} static field in {{ProcessingDetails.Timing}}
- can just use {{Timing.values()}}.
- We'll need to add documentation for new config parameters used in
{{WeightedTimeCostProvider}}.
- We'll need some tests for the patch.
> 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, HDFS-14403.002.patch, HDFS-14403.003.patch,
> HDFS-14403.004.patch, HDFS-14403.005.patch, HDFS-14403.branch-2.8.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
