You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cloudberry.apache.org by "andr-sokolov (via GitHub)" <gi...@apache.org> on 2026/05/08 12:55:05 UTC
Re: [D] [Proposal] Iceberg subsystem for datalake_fdw β design proposal [cloudberry]
GitHub user andr-sokolov added a comment to the discussion: [Proposal] Iceberg subsystem for datalake_fdw β design proposal
**Comparing the performance of Greenplum, Starrocks and Trino when reading Iceberg tables**
_Greenplum cluster for the test_
The cluster consist of 5 hosts: master, standby and 3 segment hosts. Each host has 4 CPU cores and 16GB RAM. There are 4 primaries on each segment host - one primary per CPU core. The same equipment has been used to run TPC-H queries on Trino ΠΈ StarRocks. I used TEA (https://github.com/lithium-tech/tea) to read Iceberg tables from Greenplum 6.
_Results_
<img width="1280" height="625" alt="1" src="https://github.com/user-attachments/assets/ba2a6703-22c0-409f-959f-47e87f58a930" />
The horizontal axis shows the numbers of the TPC-H test queries, and the vertical axis shows their execution time in seconds. You can find out the exact numbers in the attached [html](https://github.com/user-attachments/files/27520864/11.html). The dark red color means that the query failed.
There are `explain analyze verbose`-s for each query in [this file](https://github.com/user-attachments/files/27520800/explain.txt).
Greenplum executed only the q06 query faster, other queries are executed significantly slower.
GitHub link: https://github.com/apache/cloudberry/discussions/1683#discussioncomment-16852933
----
This is an automatically sent email for dev@cloudberry.apache.org.
To unsubscribe, please send an email to: dev-unsubscribe@cloudberry.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@cloudberry.apache.org
For additional commands, e-mail: dev-help@cloudberry.apache.org