Public Benchmark Summary

Sequential Storage

Sanitized summary of a sequential KessDB storage benchmark. The public page keeps headline performance evidence while withholding harness details and proprietary layout information.

Run Summary

38.1 MBKessDB file
1.670 sFull-fidelity save
1.447 sCompaction
0.00 BWarmed read allocations
Public Benchmark Boundary: This public summary redacts corpus identifiers, local paths, exact harness arguments, internal build identifiers, and proprietary file-format or protocol details. Full methodology and reproducibility material should be shared only under written evaluation agreement.

Public Environment

Runtime.NET NativeAOT-capable runtime
Operating systemWindows x64 workstation-class system
Architecturex64
ProcessorModern desktop-class Intel CPU
Build identifierWithheld from public summary

Dataset Summary

CorpusPrivate structured object-document corpus
Documents85,164
Input source size425,554,486 bytes
Target .kess size38,073,370 bytes
Validationclean
IndexingRead-heavy generated lookup indexes enabled

Storage Results

Hot delta save554.96 ms
Full-fidelity save1.670 s
Compaction runtime1.447 s
Storage reduction vs source JSON91.05%
Validation stateclean

Read Path Snapshot

Point read p50~32 us
Warm random point p50~1 us
Indexed equality p50~1.4 us
Steady-state warmed allocations0.00 B/read

Convenience materializers that create POCOs, strings, arrays, or lists intentionally allocate ordinary managed objects.