source: https://dipankar-tnt.medium.com/onetable-interoperability-for-apache-hudi-iceberg-delta-lake-bb8b27dd288d (coming from https://www.reddit.com/r/dataengineering/comments/18f3wo9/comment/kcry9hw/?utm_source=share&utm_medium=web2x&context=3)

the Reddit overlords seem to favour Delta though https://www.reddit.com/r/dataengineering/comments/zv2puz/which_lakehouse_table_forma[…]_you_expect_your/?utm_source=share&utm_medium=web2x&context=3

and Iceberg is the slowest https://www.onehouse.ai/blog/apache-hudi-vs-delta-lake-vs-apache-iceberg-lakehouse-feature-comparison

Here's a very Iceberg-biased article that makes the case for storage independent of execution: https://tabular.medium.com/the-case-for-independent-storage-74ac880092d9 🙂 (Tabular is an enterprise Iceberg company) I didn't get any great references (will update if i do!), but it seems a lot of Chinese companies did start with Hudi and run into challenges, and have now been migrating to Iceberg. Iceberg does see significant use in some companies like Netflix (because they created it) and Apple. Re Delta, the main issue seems to be that it was never really intended as an independent file format, and is really most focused on how it plays into the Spark ecosystem. Probably need some more unbiased benchmarks. I'm guessing some of the pro-Iceberg sentiment in the above paragraph is also more focused around streaming use cases.