[
https://issues.apache.org/jira/browse/HADOOP-17165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17222653#comment-17222653
]
Janus Chow commented on HADOOP-17165:
-------------------------------------
In HADOOP-15016, [~ywskycn] raised the idea of initializing a queue for each
special user, which seems a little complicated when the service-user list needs
frequently updated. In this ticket, the idea was to put all service-users'
request to priority level 0.
How about combining these two ideas together, to create one separate queue for
all service-users, with a "ratio" configuration specifying how much weight this
service-user queue count in all queues.
The benefits are:
# Since service-user and normal users are separate, the users used to be
assigned to priority level 0 would not be affected if a lot of service-user
requests come to level 0.
# With the configuration of the service-user queue ratio, we can have a better
throughput by setting an optimal ratio. Furthermore, the ratio can be set
dynamically, we can set the ratio to a high value when the service-user jobs
are more important and vice versa.
# Since it's a separate queue, we don't have to worry about the service-users'
cost time affecting the normal users.
> Implement service-user feature in DecayRPCScheduler
> ---------------------------------------------------
>
> Key: HADOOP-17165
> URL: https://issues.apache.org/jira/browse/HADOOP-17165
> Project: Hadoop Common
> Issue Type: New Feature
> Reporter: Takanobu Asanuma
> Assignee: Takanobu Asanuma
> Priority: Major
> Labels: pull-request-available
> Fix For: 3.4.0
>
> Attachments: HADOOP-17165.001.patch, HADOOP-17165.002.patch,
> after.png, before.png
>
> Time Spent: 2h 20m
> Remaining Estimate: 0h
>
> In our cluster, we want to use FairCallQueue to limit heavy users, but not
> want to restrict certain users who are submitting important requests. This
> jira proposes to implement the service-user feature that the user is always
> scheduled high-priority queue.
> According to HADOOP-9640, the initial concept of FCQ has this feature, but
> not implemented finally.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
