Skip to content

Pin the ravel entry to v0.21.0 - #2371

Merged
alexey-milovidov merged 3 commits into
ClickHouse:mainfrom
pmoust:ravel-v0.21.0
Oct 4, 2026
Merged

alexey-milovidov merged 3 commits into
ClickHouse:mainfrom
pmoust:ravel-v0.21.0

Conversation

@pmoust

@pmoust pmoust commented Oct 3, 2026

Copy link
Copy Markdown
Contributor

Pins the ravel entry to Ravel v0.21.0.

What changes for this benchmark:

  • Data size: the log format stores the same rows smaller, 9.8 GB instead of 11.2 GB.

  • q33: the derived per-query SQL memory pool is now the tenant's whole SQL share: 16.5 GB on c6a.4xlarge, up from 8.2 GB. The widest GROUP BY (q33, about 10.9 GB) now completes in the stock configuration instead of being refused, so the stock result covers 43 of 43 statements.

  • Cold runs: the stock fetch policy still reads whole objects, so cold time still tracks the dataset size read from the gp2 volume.

  • Measurement: the published v0.21.0 binaries, checked against the release's SHA256SUMS, on a fresh c6a.4xlarge using this repository's cloud-init with the unmodified entry:

    figure v0.19.0 v0.21.0
    load 1,248.7 s 1,234.9 s
    data size 11.24 GB 9.84 GB
    cold sum, the 42 statements v0.19.0 answered 1,712.1 s 1,252.7 s
    hot sum, same 42 (best of tries 2 and 3) about 90 s 79.8 s
    statements answered 42 of 43 43 of 43
    concurrent QPS / error ratio 0.403 to 0.502 / 0.060 to 0.091 0.482 / 0.071

The README's figures for the dataset size, the resolved pools and q33 are updated, and the earlier figures are labelled with their versions. No other change to the scripts: same flags, same load, same RustFS.

v0.21.0 is published. Its log format stores the dataset in 9.8 GB instead of
11.2 GB, and the derived per-query SQL pool is now the tenant's whole SQL
share, 16.5 GB on c6a.4xlarge instead of 8.2 GB, so q33 completes in the
stock configuration. The README's figures for the dataset size, the
resolved pools and q33 are updated, and the earlier figures are labelled
with their versions.

Signed-off-by: Panagiotis Moustafellos <pmoust@nofire.ai>
@alexey-milovidov alexey-milovidov added the machine:all PR benchmark on every machine type label Oct 4, 2026
@alexey-milovidov alexey-milovidov self-assigned this Oct 4, 2026
@alexey-milovidov
alexey-milovidov deployed to benchmark-approval October 4, 2026 03:19 — with GitHub Actions Active
@github-actions

github-actions Bot commented Oct 4, 2026

Copy link
Copy Markdown
Contributor

Results for ravel are ready for: c6a.4xlarge, c6a.metal, c7a.metal-48xl, c8g.4xlarge, c8g.metal-48xl.
The result files are committed as c1b6a8e.
The run of ravel on c6a.large did not produce results.
The run of ravel on t3a.small did not produce results.

Logs:

@github-actions

github-actions Bot commented Oct 4, 2026

Copy link
Copy Markdown
Contributor

Results for ravel are ready for: c6a.2xlarge, c6a.xlarge.
The result files are committed as d819cd2.

Logs:

@alexey-milovidov
alexey-milovidov merged commit b3759fa into ClickHouse:main Oct 4, 2026
1 of 2 checks passed

This branch is waiting to be deployed

1 waiting deployment
benchmark-approval — d819cd22 Waiting Oct 4, 2026 by robot-clickhouse via launch #645
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

machine:all PR benchmark on every machine type

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants