flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "godfrey he (JIRA)" <j...@apache.org>
Subject [jira] [Created] (FLINK-12424) Supports dag (multiple-sinks query) optimization
Date Tue, 07 May 2019 05:34:00 GMT
godfrey he created FLINK-12424:

             Summary: Supports dag (multiple-sinks query) optimization
                 Key: FLINK-12424
                 URL: https://issues.apache.org/jira/browse/FLINK-12424
             Project: Flink
          Issue Type: New Feature
          Components: Table SQL / Planner
            Reporter: godfrey he
            Assignee: godfrey he
         Attachments: image-2019-05-07-13-33-02-793.png

Currently, Flink planner will optimize the plan in {{writeToSink}} method. If there are more
than one sink in a query, each sink-tree will be optimized independent and the result execution
plans are also completely independent. Actually, there is a high probability of duplicate
computing for a multiple-sinks query. This issue aims to resolve the above problem. The basic
idea of the solution is as following: 
1. lazy optimization: does not optimize the plan in {{writeToSink}} method, just puts the
plan into a collection.
2. whole plan optimization and execution: a new {{execute}} method is added in {{TableEnvironment}},
this method will trigger whole plan optimization and execute the job.

The basic idea of dag (multiple-sinks query) optimization:
1. decompose the dag into different block, the leaf block is the common sub-plan
2. optimize each block from leaf block to root block, each block only needs to be optimized


This message was sent by Atlassian JIRA

View raw message