Skip to content

Latest commit

 

History

History
31 lines (26 loc) · 1.65 KB

solid-remediation-benchmark-report.md

File metadata and controls

31 lines (26 loc) · 1.65 KB

SOLID Remediation Benchmark Report

Date: 2026-09-09 Baseline: 5ab11d8e Profile: cntryl-stress default, three runs per row, same host

This report records the representative WebSocket regression-gate row for each domain affected by the family-ownership remediation. Throughput is the median of the three per-run throughput medians. Latency is the median of the three per-run p95 values. Queue's canonical lifecycle row does not publish a paired latency measurement.

Domain Baseline throughput Current throughput Throughput ratio Baseline p95 Current p95 p95 ratio
KV 6,888.53/s 8,279.99/s 120.2% 206.79 us 157.96 us 76.4%
Queue 7,075.09/s 7,416.73/s 104.8% n/a n/a n/a
Notice 29,356.48/s 27,389.29/s 93.3% 49.38 us 52.79 us 106.9%
RPC 12,742.87/s 14,634.62/s 114.8% 111.33 us 116.58 us 104.7%
Lease 14,780.98/s 14,955.46/s 101.2% 85.88 us 90.25 us 105.1%
Schedule 8,117.95/s 9,025.19/s 111.2% 366.50 us 247.50 us 67.5%
Stream 27,329.74/s 29,654.57/s 108.5% 60.92 us 51.17 us 84.0%

Every throughput ratio is at least 90%, and every available p95 ratio is no more than 110%. Stream also clears its stricter 95% throughput and 105% p95 limits.

During validation, broad sequential runs showed simultaneous high-variance outliers across unrelated domains. Lease and Schedule were therefore rerun by alternating the remediated tree with the immutable baseline worktree; Stream was rerun alone. The table uses those matched or isolated three-run sets rather than treating host scheduling noise as product signal.