Choose a Columnar Format From the Read Path Backward

Model projection, filtering, updates, object-store requests, and recovery before selecting a fashionable analytics file format.

lunedì 13 luglio 2026 New tab

337 words~2 min read

The columnar-format ecosystem is moving quickly. Feature tables encourage teams to ask which format has the newest encoding. A system-design decision starts elsewhere: which bytes must each workload read, how often does data change, and what happens when a writer fails?

Declare the workload

dataset:

logical_size_gib: 500

rows: 4000000000

Choose a Columnar Format From the Read Path Backward

Choose a Columnar Format From the Read Path Backward

Related reading

Using Zstd Frames to Egress Partial Parquet Files

Building on open table formats

I Built a Columnar File Format in Pure Python — a tiny, readable Parquet

The File Format Renaissance: Parquet, Lance, Vortex, Nimble, BtrBlocks, and the…

When to denormalize, when to join: A ClickHouse guide (2026)

How to align columnar output in the terminal

Related reading

Using Zstd Frames to Egress Partial Parquet Files

Building on open table formats

I Built a Columnar File Format in Pure Python — a tiny, readable Parquet

The File Format Renaissance: Parquet, Lance, Vortex, Nimble, BtrBlocks, and the…

When to denormalize, when to join: A ClickHouse guide (2026)

How to align columnar output in the terminal