mirror of
https://github.com/XRPLF/rippled.git
synced 2026-04-29 15:37:57 +00:00
f135842071
Verified CPU, memory, and network overhead calculations against official OTel C++ SDK benchmarks (969 CI runs) and source code analysis. Key corrections: - Span creation: 200-500ns → 500-1000ns (SDK BM_SpanCreation median ~1000ns; original estimate matched API no-op, not SDK path) - Per-TX overhead: 2.4μs → 4.0μs (2.0% vs 1.2%; still within 1-3%) - Active span memory: ~200 bytes → ~500-800 bytes (Span wrapper + SpanData + std::map attribute storage) - Static memory: ~456KB → ~8.3MB (BatchSpanProcessor worker thread stack ~8MB was omitted) - Total memory ceiling: ~2.3MB → ~10MB - Memory success metric target: <5MB → <10MB - AddEvent: 50-80ns → 100-200ns Added Section 3.5.4 with links to all benchmark sources. Updated presentation.md with matching corrections. High-level conclusions unchanged (1-3% CPU, negligible consensus). Also includes: review fixes, cross-document consistency improvements, additional component tracing docs (PathFinding, TxQ, Validator, etc.), context size corrections (32 → 25 bytes). Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>