Benchmarks

Reproducible. Beats every alternative.

Published ratios on realistic Kubernetes telemetry and on every public structured-JSON test corpus. Methodology is reproducible — request the bench harness and run it against your own data.

lossless · SHA-256 verified·public corpora·reproducible·beats reference
Lossless compression ratio
Logs
18.2×
Traces
27.4×
Metrics
30.1×
Sasquatchgzip
vs gzip baseline

The number every CFO can verify.

Gzip is what every observability tool ships out of the box. We ran the same corpus through both. Sasquatch wins by a factor of 8–12× on realistic Kubernetes telemetry, with full lossless round-trip preservation.

Logs
18.2×
vs gzip 2.1×

realistic K8s DaemonSet · 50K events · lossless ✓

Improvement vs gzip8.7× tighter
Traces
27.4×
vs gzip 2.4×

OTLP spans + span events · tail-sampled · lossless ✓

Improvement vs gzip11.4× tighter
Metrics
30.1×
vs gzip 2.6×

OTLP + Prom remote-write · histogram-heavy · lossless ✓

Improvement vs gzip11.6× tighter
Public corpora

Beats the leading open-source reference.

On every published structured-JSON benchmark corpus, Sasquatch produces a smaller archive than the reference open-source compressor — with full lossless round-trip on every entry.

Public corpusvs referenceround-trip
mongod105.2%lossless ✓
spark-event-log102.1%lossless ✓
cockroachdb100.4%lossless ✓
postgresql102.3%lossless ✓
Average102.5%all ✓

Numbers above 100% mean Sasquatch produces a smaller archive than the reference on the same input. Methodology and corpora are public — request the comparison harness if you want to run it yourself.

Methodology

How to reproduce these numbers on your own data.

Every benchmark in this page comes from a script you can run. No marketing-grade asterisks, no “your mileage may vary,” no test environments tuned to flatter the product.

Realistic K8s corpus

  1. 01Generate 50K events across ~20 pods with realistic K8s metadata
  2. 02Run the agent end-to-end against the corpus
  3. 03SHA-256 chain root computed on input + decompressed output
  4. 04Numbers reported are post-compaction, including overhead

Public structured-JSON corpora

  1. 01Standard public test corpora (mongod, spark-event-log, cockroachdb, postgresql)
  2. 02Run the reference open-source compressor and Sasquatch back to back
  3. 03Compare archive byte sizes and verify decompressed checksums
  4. 04Reproducer harness available on request