Hi everyone, We have published a blog article that reports the performance improvement from three patches HIVE-28488, HIVE-28489, and HIVE-28490 which we submitted some time ago. It evaluates Hive 4.0.1 on MR3 and Trino 453 on the 10TB TPC-DS benchmark, but the results could be useful to users of Apache Hive 4 (with or without LLAP).
https://www.datamonad.com/post/2024-10-09-optimizing-hive-4.0-performance/ We got the ideas for the three patches by comparing query plans generated by Hive 4 and Trino (for those queries that Trino executes much faster than Hive 4). Currently the patches are not actively reviewed, so we would appreciate it if some committers could take a look and try merging them to the master branch. If you are currently using Hive 4, backporting these patches should improve the performance on some class of queries. Thanks, --- Sungwoo Park
