Reading through this blog, to me it seems Parquet is lot like ClickHouse native data format.

Best part of ClickHouse native data format is I can use the same ClickHouse queries and can run in local or remote server/cluster and let ClickHouse to decide the available resources in the most performant way.

ClickHouse has a native and the fastest integration with Parquet so i can:

- Query local/s3 parquet data from command line using clickhouse-local.

- Query large amount of local/s3 data programmatically by offloading it to clickhouse server/cluster which can do processing in distributed fashion.