[ 
https://issues.apache.org/jira/browse/FLINK-10240?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16605399#comment-16605399
 ] 

Till Rohrmann commented on FLINK-10240:
---------------------------------------

Making Flink's scheduling mechanism more flexible and powerful is a really good 
idea. Especially on the batch side the scheduling is not always perfect. I 
would be interested if you already have concrete ideas how to improve the 
scheduling. Quite some time ago the community discussed a leg-wise scheduling 
of the {{ExecutionGraph}}. See FLINK-2119 for more details.

> Flexible scheduling strategy is needed for batch job
> ----------------------------------------------------
>
>                 Key: FLINK-10240
>                 URL: https://issues.apache.org/jira/browse/FLINK-10240
>             Project: Flink
>          Issue Type: New Feature
>          Components: Distributed Coordination
>            Reporter: Zhu Zhu
>            Priority: Major
>              Labels: scheduling
>
> Currently in Flink we have 2 schedule mode:
> 1. EAGER mode starts all tasks at once, mainly for streaming job
> 2. LAZY_FROM_SOURCES mode starts a task once its input data is consumable
>  
> However, in batch job, input data ready does not always mean the task can 
> work at once. 
> One example is the hash join operation, where the operator first consumes one 
> side(we call it build side) to setup a table, then consumes the other side(we 
> call it probe side) to do the real join work. If the probe side is started 
> early, it just get stuck on back pressure as the join operator will not 
> consume data from it before the building stage is done, causing a waste of 
> resources.
> If we have the probe side task started after the build stage is done, both 
> the build and probe side can have more computing resources as they are 
> staggered.
>  
> That's why we think a flexible scheduling strategy is needed, allowing job 
> owners to customize the vertex schedule order and constraints. Better 
> resource utilization usually means better performance.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to