Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

perf(over window): shortcut implementation for first_value and last_value in over window #19233

Merged
merged 1 commit into from
Nov 1, 2024

Conversation

stdrc
Copy link
Member

@stdrc stdrc commented Nov 1, 2024

I hereby agree to the terms of the RisingWave Labs, Inc. Contributor License Agreement.

What's changed and what's your intention?

Previously we reuse batch aggregator for first_value and last_value in OverWindow, and this required an O(n) iteration through the "current window" when evaluating output for each row. This PR introduces shortcut implementations fro these two special aggregate types to avoid the need of O(n) iteration.

This also benefits lag and lead because they are implemented with first_value and last_value under the hood.

A rough benchmark with a dumb query based on nexmark source showed a 392% source throughput increase introduced by this PR, during the first 10 minutes after streaming job creation.

Query:

CREATE VIEW all_points AS
SELECT
    account,
    key,
    SUM(delta) AS points
FROM (
    SELECT
        ((bidder * 2654435761) # (bidder >> 16)) % 10000 as account,
        auction % 3 as key,
        price % 10 as delta
    FROM bid
)
GROUP BY account, key;

-- rank = row_number + first_value

set rw_streaming_over_window_cache_policy = full;
CREATE MATERIALIZED VIEW rn AS
SELECT
    account,
    key,
    points,
    ROW_NUMBER() OVER (
        PARTITION BY key
        ORDER BY points DESC
    ) AS rn
FROM all_points;

set rw_streaming_over_window_cache_policy = recent;
CREATE SINK mv AS
SELECT
    account,
    key,
    points,
    FIRST_VALUE(rn) OVER (
        PARTITION BY key
        ORDER BY points DESC
        RANGE BETWEEN CURRENT ROW AND CURRENT ROW
    ) AS rank
FROM rn
WITH ( connector = 'blackhole', type = 'append-only', force_append_only = 'true');

Before:

before

After:

after

Checklist

  • I have written necessary rustdoc comments
  • I have added necessary unit tests and integration tests
  • I have added test labels as necessary. See details.
  • I have added fuzzing tests or opened an issue to track them. (Optional, recommended for new SQL features Sqlsmith: Sql feature generation #7934).
  • My PR contains breaking changes. (If it deprecates some features, please create a tracking issue to remove them in the future).
  • All checks passed in ./risedev check (or alias, ./risedev c)
  • My PR changes performance-critical code. (Please run macro/micro-benchmarks and show the results.)
  • My PR contains critical fixes that are necessary to be merged into the latest release. (Please check out the details)

Documentation

  • My PR needs documentation updates. (Please use the Release note section below to summarize the impact on users)

Release note

If this PR includes changes that directly affect users or other significant modifications relevant to the community, kindly draft a release note to provide a concise summary of these changes. Please prioritize highlighting the impact these changes will have on users.

Copy link
Member Author

stdrc commented Nov 1, 2024

This stack of pull requests is managed by Graphite. Learn more about stacking.

Join @stdrc and the rest of your teammates on Graphite Graphite

@stdrc stdrc changed the title add shortcut for first_value and last_value in over window impl perf(over window): shortcut implementation for first_value and last_value in over window Nov 1, 2024
@stdrc stdrc marked this pull request as ready for review November 1, 2024 09:24
Copy link
Member

@BugenZhao BugenZhao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Really so 🐮?

Copy link
Member

@BugenZhao BugenZhao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Really so 🐮?

Copy link
Member

@BugenZhao BugenZhao left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Really so 🐮?

@stdrc stdrc added this pull request to the merge queue Nov 1, 2024
Merged via the queue into main with commit b39c9af Nov 1, 2024
51 of 53 checks passed
@stdrc stdrc deleted the rc/over_window_first_last_value_shortcut branch November 1, 2024 10:47
stdrc added a commit that referenced this pull request Nov 4, 2024
…t_value` in over window (#19233)

Signed-off-by: Richard Chien <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants