feature-benchmark: Add MaterializedViewSink #30762
Draft
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
If the append operator discards a batch because it can already not be appended anymore, we should call `delete()` on the batch before dropping it. This doesn't actually change any behavior, since batch deletion is disable in production currently, but it avoids triggering the WARN log in `Batch::drop`.
This commit makes the MV sink always perform a consolidation of (the relevant parts of) the correction buffer after it has received all updates for the snapshot, both from `desired` and `persist`. We know that these updates will cancel out, so we have an opportunity to shed a large amount of memory by forcing a consolidation at this time. This is especially important in read-only mode when the MV sink does not produce any batches, which is the process that normally forces consolidation to happen.
teskje
reviewed
Dec 7, 2024
misc/python/materialize/feature_benchmark/scenarios/benchmark_main.py
Outdated
Show resolved
Hide resolved
misc/python/materialize/feature_benchmark/scenarios/benchmark_main.py
Outdated
Show resolved
Hide resolved
Comment on lines
+1582
to
+1583
| > CREATE MATERIALIZED VIEW sink1_check_v AS SELECT COUNT(*) FROM sink1_check_tbl; | ||
|
|
||
| > SELECT * FROM sink1_check_v |
There was a problem hiding this comment.
If this is just to have a way to know when the sink has finished writing the data, and not to test MV performance, is there a reason to create the sink1_check_v? Couldn't we just SELECT from the sink1_check_tbl directly instead?
There was a problem hiding this comment.
Then it's what the ExactlyOnce scenario is already doing.
Local run: NAME | TYPE | THIS | OTHER | UNIT | THRESHOLD | Regression? | 'THIS' is -------------------------------------------------------------------------------------------------------------------------------------------------------- MaterializedViewSink | wallclock | 11.270 | 31.620 | s | 10% | no | better: 2.8 times faster MaterializedViewSink | memory_mz | 3699.303 | 3870.010 | MB | 20% | no | better: 4.4% less MaterializedViewSink | memory_clusterd | 258.636 | 294.971 | MB | 50% | no | better: 12.3% less
def-
force-pushed
the
pr-test-mv-sink-v2
branch
from
b8525d2 to
5038023
Compare
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.
@teskje Does this look reasonable for you? Only look at the last commit, it's based on #30758
Checklist
$T ⇔ Proto$Tmapping (possibly in a backwards-incompatible way), then it is tagged with aT-protolabel.