CargoShip

Appearance

Benchmarking ​

cargoship benchmark measures how different compression algorithms and levels perform on your data, so you can trade archive size against CPU time before deciding whether to override the automatic level with --compression-level on real uploads. It reports compression ratio and throughput for gzip, zlib, zstd, lz4, and s2.

bash
# Benchmark against generated data of a given size and type
cargoship benchmark --size 100MB --data-type text

# Benchmark against a real file (most representative)
cargoship benchmark --file /path/to/sample.tar

# Emit machine-readable results
cargoship benchmark --size 1GB --format json

Options ​

  • --size — size of generated test data, e.g. 1MB, 10MB, 1GB (default 10MB). Ignored when --file is set.
  • --data-type — shape of generated data: text, binary, mixed (default), or random. Text is highly compressible; random is not, which brackets the range you'll see in practice.
  • --file — benchmark a real file instead of generated data. Use this for the most representative result on your workload.
  • --format — table (default, human-readable) or json for scripting.

Reading the results ​

Table output lists each algorithm and level with the ratio it achieved and its throughput. A run on compressible text data looks roughly like this:

Algorithm  Level  Ratio   Compress MB/s   Decompress MB/s
gzip       6      3.1x    95              420
zlib       6      3.1x    98              430
zstd       3      3.4x    480             1250
zstd       19     3.9x    24              1180
lz4        -      2.4x    720             2600
s2         -      2.6x    900             2400

Higher levels shrink the archive further but cost CPU: notice zstd at level 19 edges out level 3 on ratio while running an order of magnitude slower to compress. lz4 and s2 prioritize speed over ratio. Your numbers will vary with data type and hardware — always benchmark your own --file before picking a level for large jobs. CargoShip uploads use Zstandard; the benchmark's cross-algorithm view helps you judge how much a higher --compression-level is worth for your content.

See also ​