Skip to content

Add ZigHouse#895

Merged
alexey-milovidov merged 2 commits into
ClickHouse:mainfrom
donge:add-zighouse-release
May 12, 2026
Merged

Add ZigHouse#895
alexey-milovidov merged 2 commits into
ClickHouse:mainfrom
donge:add-zighouse-release

Conversation

@donge
Copy link
Copy Markdown
Contributor

@donge donge commented May 11, 2026

Summary

  • Add ZigHouse as a ClickBench entry using a fixed Linux x86_64 release binary.
  • Include runner scripts compatible with ClickBench's shared benchmark driver.
  • Include AWS c6i.4xlarge results from the 2026-05-11 run.

Notes

The result uses c6i.4xlarge because c6a.4xlarge was unavailable in the AWS China regions used for this run.

Release binary:
https://github.com/donge/zighouse/releases/tag/v0.1.0-clickbench

SHA256:
5a779eacf87082eeeb13b336d6f798a0399593c29b5a586ab67202104396dc83

@rschu1ze
Copy link
Copy Markdown
Member

Is there a homepage or GitHub repository for ZigHouse? #836 made me a bit sceptical about submissions that only come with a download but not homepage/repository.

@alexey-milovidov
Copy link
Copy Markdown
Member

alexey-milovidov commented May 11, 2026

I'm in favor of accepting everything that works at least occasionally. I like looking at experimental, obscure, discontinued, historical, and even crackpot databases: https://presentations.clickhouse.com/2018-unknown_developers_reissue/

Just take care when testing - use only isolated machines.
We can do forensic analysis as well :)

@donge
Copy link
Copy Markdown
Contributor Author

donge commented May 11, 2026

Thanks for your feedback. I will try to make the submission satisfy the project requirements as much as possible.
ZigHouse is still experimental at this stage, so let me briefly explain what it is.

I have been a long-time ClickHouse user, and I learned a lot from ClickHouse's design while using it. ZigHouse is an attempt to build a lightweight single-node analytical database inspired by column-oriented execution, vectorized data processing, and SIMD-friendly kernels.

I also used DuckDB as an important reference point when comparing the design and behavior. At the current stage, ZigHouse is closer to a non-general-purpose, embedded DuckDB-like analytical engine rather than a mature database system.

The main target use case is lightweight embedded analytics on structured wide tables, especially time-related analytical workloads with scans, aggregations, Top-K queries, and unique/distinct-style computations.
The repository is here:
https://github.com/donge/zighouse

The slides looks fun, let me check and update the progress, Thank you all

@donge
Copy link
Copy Markdown
Contributor Author

donge commented May 12, 2026

Here is a simple homepage https://donge.github.io/zighouse/

@alexey-milovidov alexey-milovidov self-assigned this May 12, 2026
@alexey-milovidov
Copy link
Copy Markdown
Member

The scripts work.

@alexey-milovidov alexey-milovidov merged commit d3b341c into ClickHouse:main May 12, 2026
@alexey-milovidov
Copy link
Copy Markdown
Member

Added a result here: #898

@ClickHouse ClickHouse deleted a comment from CLAassistant May 12, 2026
@donge
Copy link
Copy Markdown
Contributor Author

donge commented May 13, 2026

Thanks Alexey, truly impressive with your productivity as CTO coder

@alexey-milovidov
Copy link
Copy Markdown
Member

ZigHouse hardcodes the set of ClickBench queries and can't process any other query, so we will have to remove it.

@donge
Copy link
Copy Markdown
Contributor Author

donge commented May 14, 2026

Sure, ZigHouse is still being built bottom-up right now.

My goal is to set the fastest possible baseline so we don't see any performance drops when integrating generic path later on. this time only have an executor for now, without a complete SQL AST parser.
I will make it work with common SQL next time. anyway, thanks for providing a project like this—it sets a great common benchmark.

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

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants