[ 
https://issues.apache.org/jira/browse/FLINK-11946?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Jing Zhang updated FLINK-11946:
-------------------------------
    Description: 
This issue introduces ExecNode. ExecNode and FlinkPhysicalRel plays different 
role. FlinkPhysicalRel is a physical relational expression, while ExecNode is a 
representation of execution information, such as how much resource the node 
will take, how to convert to StreamTransformation. 
It should be noticed that FlinkPhysicalRel and ExecNode could be implemented by 
a same concrete physical node, such as StreamExecCalc, StreamExecScan, etc. The 
interface exposed by a concrete physical node is different at different stages. 
A FlinkPhysicalRel DAG is generated after physical optimization, then translate 
the plan to ExecNode DAG. All post processors all apply on ExecNode DAG, such 
as detect deadlock, set heap/memory/cpu, etc.

  was:
This issue introduces Exec nodes. Each exec node contains execution 
informations, such as how much resource the node will take, how to convert to  
StreamTransformation. 

 


> Introduce ExecNode
> ------------------
>
>                 Key: FLINK-11946
>                 URL: https://issues.apache.org/jira/browse/FLINK-11946
>             Project: Flink
>          Issue Type: Task
>          Components: API / Table SQL
>            Reporter: Jing Zhang
>            Priority: Major
>
> This issue introduces ExecNode. ExecNode and FlinkPhysicalRel plays different 
> role. FlinkPhysicalRel is a physical relational expression, while ExecNode is 
> a representation of execution information, such as how much resource the node 
> will take, how to convert to StreamTransformation. 
> It should be noticed that FlinkPhysicalRel and ExecNode could be implemented 
> by a same concrete physical node, such as StreamExecCalc, StreamExecScan, 
> etc. The interface exposed by a concrete physical node is different at 
> different stages. A FlinkPhysicalRel DAG is generated after physical 
> optimization, then translate the plan to ExecNode DAG. All post processors 
> all apply on ExecNode DAG, such as detect deadlock, set heap/memory/cpu, etc.



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

Reply via email to