You are viewing a plain text version of this content. The canonical link for it is here.
Posted to yarn-issues@hadoop.apache.org by "Yufei Gu (JIRA)" <ji...@apache.org> on 2017/10/15 09:55:02 UTC
[jira] [Commented] (YARN-6460) Accelerated time in SLS
[ https://issues.apache.org/jira/browse/YARN-6460?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16205084#comment-16205084 ]
Yufei Gu commented on YARN-6460:
--------------------------------
Thanks [~seneque] for working on this. The approach looks good to me generally. The patch doesn't apply and needs a rebase though.
> Accelerated time in SLS
> -----------------------
>
> Key: YARN-6460
> URL: https://issues.apache.org/jira/browse/YARN-6460
> Project: Hadoop YARN
> Issue Type: Improvement
> Components: scheduler-load-simulator
> Reporter: Julien Vaudour
> Assignee: Julien Vaudour
> Priority: Minor
> Attachments: YARN-6460-branch-2.000.patch, YARN-6460-branch-2.001.patch, YARN-6460-branch-2.002.patch, YARN-6460.000.patch, YARN-6460.001.patch, YARN-6460.002.patch
>
>
> Be able to accelerate time in SLS.To do that, a {{timescalefactor}} parameter is introduced (default value = 1)
> With if we use a time factor of X, time in simulation will be X times faster than real. Time in generated CSV will be modified a well to have the same result as if we don't use {{timescalefactor}}.
> For example, it permits to run a simulation of one week of jobs in just one day if we use {{timescalefactor=7}}
> To do that a ScaleClock object has been introduced, which implements {{org.apache.hadoop.yarn.util.Clock}}. It also extends {{com.codahale.metrics.Clock}} for metrics reported in csv.
> All objects used for the simulation now use a shared reference to a {{org.apache.hadoop.yarn.util.Clock}} instance to get current time instead of using {{System.currentTimeMillis()}}
> A new optionnal parameter {{--timescalefactor=<factor>}} has been introduced on {{slsrun.sh}} script.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org