Benchmarks

Performance benchmarks comparing Jetliner with other Avro readers.

Coming Soon

Detailed benchmark results will be added in a future update.

Benchmark Setup

Benchmarks are run using:

• Hardware: Apple M-series / Intel Xeon
• Files: Various sizes from 1MB to 10GB
• Codecs: null, snappy, zstd
• Metrics: Throughput (MB/s), memory usage, latency

Running Benchmarks

To run benchmarks locally:

# Rust benchmarks
poe bench

# Read throughput benchmark
poe bench-read

Preliminary Results

Reader	Throughput	Memory	Notes
Jetliner	TBD	TBD	Streaming, query optimization
fastavro	TBD	TBD	Pure Python
Apache Avro	TBD	TBD	Reference implementation

Factors Affecting Performance

1. Codec: Decompression is often the bottleneck
2. Schema complexity: Nested types are slower
3. Column count: More columns = more work
4. Network latency: For S3, network dominates
5. Buffer configuration: Affects prefetch efficiency