Benchmarks

This page summarizes the benchmark material that helps evaluate HestiaStore. Use it to understand workload shape and relative trade-offs before going deeper into architecture or tuning.

What the benchmark pages are for

• Compare HestiaStore against alternative embedded engines under a consistent local benchmark harness.
• Show workload direction, not absolute guarantees.
• Provide enough methodology to judge whether a result is relevant for your production workload.

Write throughput

What this view is useful for:

• estimating ingestion headroom for append-heavy local workloads
• comparing compression and filtering overhead against raw write throughput
• spotting whether your expected workload is closer to log-structured or low-latency point-write engines

Detailed page: Write Throughput Results

Random read throughput

What this view is useful for:

• estimating point-lookup behavior on pre-populated datasets
• comparing negative-lookup friendly designs with disk-backed structures
• understanding where HestiaStore fits relative to memory-mapped or native storage engines

Detailed page: Random Read Results

Sequential read throughput

What this view is useful for:

• evaluating scan-heavy or export-style workloads
• comparing sparse-index and iterator performance across engines
• checking whether your main pressure is scan throughput instead of random access latency

Detailed page: Sequential Read Results

Methodology summary

• All benchmark suites run with the same JVM, hardware, and dataset shape for a given comparison.
• Benchmarks use warm-up plus sustained measurement windows; treat the results as relative comparisons, not vendor-style peak numbers.
• Storage size, CPU use, and confidence intervals matter as much as raw throughput.
• If your workload mixes scans, hot reads, WAL durability, and background maintenance, validate with your own representative benchmark before drawing conclusions.

How to read the results

• Prefer the chart direction and rank ordering over small score deltas.
• Use the Alternatives page to turn benchmark results into a fit decision.
• Use Performance Model & Sizing if HestiaStore is already selected and you need tuning guidance.

Detailed results and raw artifacts

Detailed benchmark pages stay inside the Evaluate section so they can grow as new benchmark suites are added:

• Write Throughput Results
• Random Read Results
• Sequential Read Results

Raw benchmark artifacts and compare tooling remain in the repository:

• Benchmarks module README
• Benchmark history and comparisons