> To enhance query execution speed, we embedded DuckDB as the execution engine for columnstore queries So is it Postgres or DuckDB that cracked the analytics top ?
DuckDB
Postgres Just Cracked the Top Fastest Databases for Analytics
11–20 of 126 posts
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#12How is this different from Crunchy Warehouse which is also built on Postgres and DuckDB? https://www.crunchydata.com/products/warehouse
Think that use pure parquet, not iceberg tables.
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#13How is this different from Crunchy Warehouse which is also built on Postgres and DuckDB? https://www.crunchydata.com/products/warehouse
Architecturally, it's very similar. We have the same vision! And they've done a great job, especially around writing to Iceberg. Some differences:
1. Our extension is fully open-source. I believe they've open-sourced bits of their stack. 2. We are unopionated about open table formats: Iceberg and Delta. 3. In v0.2, we will support small write workloads. This will open up time-series and HTAP workloads.
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#14Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#15> To enhance query execution speed, we embedded DuckDB as the execution engine for columnstore queries So is it Postgres or DuckDB that cracked the analytics top ?
well, pg_mooncake is a Postgres extension, and Postgres + pg_mooncake is still just Postgres. Users deploy pg_mooncake as a Postgres extension and write and query all tables through psql.
Fast analytic databases need two key things: columnar storage and a vectorized execution engine. We introduce a columnstore table access method in Postgres with data stored in Parquet) and execute queries on those tables using DuckDB.
By leveraging DuckDB's execution engine, we avoid reinventing vectorized query execution while keeping everything managed through Postgres.
More on our architecture: https://www.mooncake.dev/blog/how-we-built-pgmooncake
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#16Looks like they're currently No. 12 on their cited Clickbench benchmark - so not quite in the top 10 three weeks later.
I know! It's crazy how quickly everyone is moving in this space.
There are some still low-hanging fruit optimizations we're working on (https://github.com/Mooncake-Labs/pg_mooncake/issues/82), and we'd love to get back to the Top 10.
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#17Looking at the effort being put in it seems there’s a massive market for proper Postgres analytics / columnstore, which is very telling of the pain involved in moving data between systems.
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#18It is not really Postgres, the queries are run on DuckDB. Yeah, DuckDB is really fast for analytical queries.
Well, pg_mooncake is a Postgres extension. So it's really just Postgres :)
We embed DuckDB in our extension as the vectorized execution engine on columnstore tables.
Why rebuild a vectorized execution, when DuckDB has a lot of GREAT properties for this workload (similar syntax to Postgres, embedability).
Here's our architecture: https://www.mooncake.dev/blog/how-we-built-pgmooncake
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#19All you need is postgres: part 73
Re: Postgres Just Cracked the Top Fastest Databases for Analytics
#20It is not really Postgres, the queries are run on DuckDB. Yeah, DuckDB is really fast for analytical queries.
oh and I totally forgot. pg_mooncake is actually faster than DuckDB on parquet.
We implement segment-elimination to do so. We even blogged about it: https://www.mooncake.dev/blog/duckdb-parquet