[
https://issues.apache.org/jira/browse/YARN-6361?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16129592#comment-16129592
]
Yufei Gu commented on YARN-6361:
--------------------------------
Hi [~daemon], thanks for the patch. Let's talk about the idea first.
{quote}
Avoid frequent ordering, and it's pointless and a waste of time to do a
sequence before each assign container. Because, after each assignment, the
whole child nodes of the queue are basically staying in order.
{quote}
Agreed basically, but it's the right thing to do until we find a solution to
either avoid sorting every time and guarantee fairness between queues.
{quote}
And we don't really need to ensure that all of our fair shares is guaranteed,
after all, even though we do a sort of order before each of the container's
assignment because the FSQueue#demand is updated in the last time the
FairScheduler# update cycle.
So the value of demand is not real time, which also leads to the fact that we
are not strictly and fairly shared.
So, we can sort all the queues at the FairScheduler#update cycle, and we now
have a default of 0.5 s per update cycle, which is worth doing.
{quote}
We should ensure fairness as possible as we can. Demand doesn't play a big role
in {{FairShareComparator}}. It only matters while resource usage is less than
min share. In addition, applications like MR don't change their demands
frequently. In most cases, there are only 2 times for a job, which is asking
resources for all Mapper and all Reducer.
The community is working on the global scheduling, which will mitigate this
performance issue dramatically in different ways.
BTW, your second point is legit. YARN-4090 is quiet for a while, you are
welcome to contribute if you want.
> FairScheduler: FSLeafQueue.fetchAppsWithDemand CPU usage is high with big
> queues
> --------------------------------------------------------------------------------
>
> Key: YARN-6361
> URL: https://issues.apache.org/jira/browse/YARN-6361
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Miklos Szegedi
> Assignee: YunFan Zhou
> Attachments: dispatcherthread.png, threads.png, YARN-6361.001.patch
>
>
> FSLeafQueue.fetchAppsWithDemand sorts the applications by the current policy.
> Most of the time is spent in FairShareComparator.compare. We could improve
> this by doing the calculations outside the sort loop {{(O\(n\))}} and we
> sorted by a fixed number inside instead {{O(n*log\(n\))}}. This could be an
> performance issue when there are huge number of applications in a single
> queue. The attachments shows the performance impact when there are 10k
> applications in one queue.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
