Skip to content

Update ClickBench benchmarks with DataFusion 54.0.0 (when released) #21505

Closed

Description

@alamb

Is your feature request related to a problem or challenge?

Describe the solution you'd like

Once DataFusion 54.0.0 is released, It would be great to update ClickBench https://benchmark.clickhouse.com/ with the newest version and highlight and improvements.

ClickBench is a benchmark heavy on filtering and aggregation that we have used as an optimization target for the last several releases.

Describe alternatives you've considered

No response

Additional context

No response

Activity

  1. self-assigned this
    on Jun 1, 2026
  2. alamb commented on Jun 1, 2026

    @alamb
    ContributorAuthor

    I have been working on this. The ClickBench scripts have been refactored and I hit a few issues gathering info

    However, now I think I have enough to regenerate the results

  3. alamb commented on Jun 1, 2026

    @alamb
    ContributorAuthor

    Here are the current results (still not quite at DuckDB) but close
    Image

    Here is the url I used locally

    Here is the full result:

    ClickBench-branch-54.zip

  4. alamb commented on Jun 18, 2026

    @alamb
    ContributorAuthor

    I am starting to work on this

    Currently ClickBench has reports for these combinations:

    ┌────────────────┬───────────────────┬────────────────────────────────────────┐
    │ Instance type  │    Arch (AMI)     │                 Notes                  │
    ├────────────────┼───────────────────┼────────────────────────────────────────┤
    │ c6a.xlarge     │ x86_64 (AMD)      │ 4 vCPU                                 │
    ├────────────────┼───────────────────┼────────────────────────────────────────┤
    │ c6a.2xlarge    │ x86_64 (AMD)      │ 8 vCPU                                 │
    ├────────────────┼───────────────────┼────────────────────────────────────────┤
    │ c6a.4xlarge    │ x86_64 (AMD)      │ 16 vCPU — the canonical/reference size │
    ├────────────────┼───────────────────┼────────────────────────────────────────┤
    │ c6a.metal      │ x86_64 (AMD)      │ bare metal                             │
    ├────────────────┼───────────────────┼────────────────────────────────────────┤
    │ c7a.metal-48xl │ x86_64 (AMD)      │ bare metal, 192 vCPU                   │
    ├────────────────┼───────────────────┼────────────────────────────────────────┤
    │ c8g.4xlarge    │ arm64 (Graviton4) │ 16 vCPU                                │
    ├────────────────┼───────────────────┼────────────────────────────────────────┤
    │ c8g.metal-48xl │ arm64 (Graviton4) │ bare metal, 192 vCPU                   │
    └────────────────┴───────────────────┴────────────────────────────────────────┘
    

    I am working to get some numbers

  5. alamb commented on Jun 19, 2026

    @alamb
    ContributorAuthor

    Preliminary Results from

    We are technically faster than DuckDB at the very highest end:
    Image

    Though for the classic c6a.4xlarge we are still a bit slower
    Image

  6. alamb commented on Jun 19, 2026

    @alamb
    ContributorAuthor

    🤔 maybe I can write a blog post about this anyways (and highlight the differences in the hardware 🤔 )

  7. alamb commented on Jun 19, 2026

    @alamb
    ContributorAuthor
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions