Best Practices for Normalizing Large Datasets

I'm working on a dataset with over 10 million rows and I'm running into performance issues after normalizing to 4NF. Has anyone tackled similar scale? What indexing strategies or partitioning approaches did you find most effective?

For massive tables, I usually go with hash partitioning on the most selective column, then add a covering index on the foreign keys used most often. Also, consider using columnstore indexes if you need heavy analytical queries.

I’d also recommend looking into sharding if your RDBMS supports it. In our case, splitting the data across three nodes cut query latency by ~40%. Just be careful with cross‑shard joins.