perf: attribute pending-backfill startup (1.91s ready, 3.57s usable p50) #1225

Open
opened 2026-10-06 19:33:17 +00:00 by kayg · 0 comments
Owner

Follow-up from #1124 and #1161. Round 2 measured the fixed release on the perf VM through perf-run and the 8 ms / 200 IOPS HDD scope. Same real Home: 5,000 Files, 240 asserted Log rows, 121 asserted provider Events. Five same-release starts with the existing Derived data completion markers cleared while stopped:

Clock p50 / p95 / max (ms)
Process start → ready 1,906.53 / 2,920.83 / 2,920.83
Process start → usable Files row 3,568.52 / 4,396.51 / 4,396.51

Server RSS at usable paint: 138,760,192–141,856,768 bytes. Post-ready CPU: 22–36 ticks. One-minute load: 5.92–6.72; slow samples are retained. No matching baseline.json fixture exists. No startup change was made.

Evidence: docs/perf/2026-10-06-1124-round2.md and runs/2026-10-06-1124-round2-backfill-and-traces.json. This is backfill replay on one release, not an actual cross-release upgrade. Use the larger #1161 fixture and a frozen prior/current server pair to attribute the ready delay. Preserve stable source identities and resumable backfills. Measure the same clocks interleaved; do not hide backfill time behind a readiness-only result.

Follow-up from #1124 and #1161. Round 2 measured the fixed release on the perf VM through perf-run and the 8 ms / 200 IOPS HDD scope. Same real Home: 5,000 Files, 240 asserted Log rows, 121 asserted provider Events. Five same-release starts with the existing Derived data completion markers cleared while stopped: | Clock | p50 / p95 / max (ms) | | --- | ---: | | Process start → ready | 1,906.53 / 2,920.83 / 2,920.83 | | Process start → usable Files row | 3,568.52 / 4,396.51 / 4,396.51 | Server RSS at usable paint: 138,760,192–141,856,768 bytes. Post-ready CPU: 22–36 ticks. One-minute load: 5.92–6.72; slow samples are retained. No matching baseline.json fixture exists. No startup change was made. Evidence: docs/perf/2026-10-06-1124-round2.md and runs/2026-10-06-1124-round2-backfill-and-traces.json. This is backfill replay on one release, not an actual cross-release upgrade. Use the larger #1161 fixture and a frozen prior/current server pair to attribute the ready delay. Preserve stable source identities and resumable backfills. Measure the same clocks interleaved; do not hide backfill time behind a readiness-only result.
Commenting is not possible because the repository is archived.
No labels
No milestone
No project
No assignees
1 participant
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/calternal-v0#1225
No description provided.