Yes, the three engines used the same Metastore instance (of Hive 4.2.0), reading the same HDFS data in the same cluster. (I can test Hive-LLAP 4.3.0 in the same cluster in the future.) Thanks, --- Sungwoo
On Wed, Aug 26, 2026 at 8:19 PM Ayush Saxena <[email protected]> wrote: > Thanx Sungwoo, I pushed this to dev@ for the developers to have a view > on this. Overall very impressive, Just a question was the the > FileSystem and the hardware same for all the Query Engines? > > I think maybe 4.3.0-SNAPSHOT current master would be having better > numbers, there were some efforts going on around this since months, > nothing near this but. Nice Work!!! > > -Ayush > > > On Fri, 7 Aug 2026 at 14:34, Sungwoo Park <[email protected]> wrote: > > > > Hi everyone, > > We published a blog that tests Hive 4.2.0 LLAP, Trino 483 (latest > version), Hive 4.2.0 on MR3. > > It uses the TPC-DS benchmark of 10TB. > > Hopefully you find it useful. > > > > https://mr3docs.datamonad.com/blog/2026-08-02-performance-evaluation-3.0 > > > > --- Sungwoo >
