[Coral-Incremental] Cost calculation for RelNode #516
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
yyy1000
changed the title
yyy1000
commented
Jul 10, 2024
Comment on lines
61
to
72
| public CostInfo getExecutionCost(RelNode rel) { | ||
| if (rel instanceof TableScan) { | ||
| return getExecutionCostTableScan((TableScan) rel); | ||
| } else if (rel instanceof LogicalJoin) { | ||
| return getExecutionCostJoin((LogicalJoin) rel); | ||
| } else if (rel instanceof LogicalUnion) { | ||
| return getExecutionCostUnion((LogicalUnion) rel); | ||
| } else if (rel instanceof LogicalProject) { | ||
| return getExecutionCostProject((LogicalProject) rel); | ||
| } | ||
| return new CostInfo(0.0, 0.0); | ||
| } |
There was a problem hiding this comment.
This is the core cost calculation framework.
It
- Walk through the whole RelNode tree and get the cost
- Keep the Row Count Information which will be used for IO cost
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Outdated
Show resolved
Hide resolved
| } | ||
| } | ||
|
|
||
| private Double estimateJoinSelectivity(List<JoinKey> joinKeys) { |
There was a problem hiding this comment.
I didn't write the logic now,
it will be from the join key, and from the statistic map info like distinct/row_count to get Selectivity
yyy1000
changed the title
yyy1000
changed the title
yyy1000
changed the title
KevinGe00
requested changes
Jul 26, 2024
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Outdated
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Outdated
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Outdated
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Outdated
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Outdated
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Outdated
Show resolved
Hide resolved
| private CostInfo getExecutionCostJoin(LogicalJoin join) { | ||
| RelNode left = join.getLeft(); | ||
| RelNode right = join.getRight(); | ||
| if (!(left instanceof TableScan) || !(right instanceof TableScan)) { |
There was a problem hiding this comment.
Are expanding the cost estimator to handle joins for more than 2 tables?
Co-authored-by: Kevin Ge <kevingehd@gmail.com>
yyy1000
changed the title
KevinGe00
reviewed
Jul 29, 2024
coral-incremental/src/main/java/com/linkedin/coral/incremental/RelNodeCostEstimator.java
Show resolved
Hide resolved
coral-incremental/src/test/java/com/linkedin/coral/incremental/RelNodeCostEstimatorTest.java
Show resolved
Hide resolved
coral-incremental/src/main/java/com/linkedin/coral/incremental/TableStatistic.java
Outdated
Show resolved
Hide resolved
|
Overall feedback is that I believe the PR sounds too specific and there is no general framework for cost estimation that is applicable to various operators by specializing some of its aspects. I think we need to iterate over this PR to make it more abstract with various implementations. This may be better discussed over a doc. |
wmoustafa
reviewed
Aug 2, 2024
| // The number of rows in the table | ||
| Double rowCount; | ||
| // The number of distinct values in each column | ||
| Map<String, Double> distinctCountByRow; |
There was a problem hiding this comment.
Does this work for nested columns? If yes, what is the spec of the key?
Does it work for complex types?
If the answer is no, we should be explicit about this at least in the Java doc.
Comment on lines
194
to
195
| // The shuffle cost of a join is the maximum shuffle cost of its children because | ||
| // in modern distributed systems, the shuffle cost is dominated by the largest shuffle. |
There was a problem hiding this comment.
Terminology is not clear here. What is the shuffle cost of children? We might need to discuss this over a doc.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What changes are proposed in this pull request, and why are they necessary?
The goal of this PR is to introduce cost calculation method for Coral Incremental module, that's given an Input RelNode, return its cost for {executing it and writing the result into output}. So, for two RelNodes, one is incremental format, and the other is batch format, we could choose one with less cost, which could have better performance when executing.
The main changes are:
How was this patch tested?
Unit test