nathan/vctp2
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update ldap
continuous-integration/drone/push Build is passing
Details

Browse Source
This commit is contained in:
Nathan Coad
2026-04-21 11:00:40 +10:00
parent 361ba7719b
commit 4b1b985862
4 changed files with 174 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files
plan.md
+9 -4
View File
@@ -305,10 +305,15 @@ The target architecture is:
- [x] Validate/add canonical `vm_hourly_stats` indexes for snapshot time, vCenter+time, VM identity+time, and trace lookup.
- [x] Add PostgreSQL monthly partitioning for `vm_hourly_stats` behind migration controls.
- [x] Benchmark Go vs SQL on canonical Postgres tables using representative production-scale data.
- Production-scale Postgres run completed on 2026-04-21 via one-shot canonical benchmark (`-benchmark-aggregations` with `runs_per_mode=1`, `driver=postgres`).
- Daily window `2026-04-20T00:00:00Z` to `2026-04-21T00:00:00Z`: Go `4.000602432s` (`14881` rows) vs SQL `1h17m19.039092561s` (`14920` rows), with Go ~`1159.59x` faster on this run.
- Monthly window `2026-04-01T00:00:00Z` to `2026-05-01T00:00:00Z`: Go `3.529410947s` (`15871` rows) vs SQL `3.313037973s` (`15873` rows), near parity with SQL slightly faster (~`0.216s`, `6.1%`).
- Decision remains unchanged: keep Go as scheduled default and treat SQL as fallback/backfill until SQL shows a clear, repeatable runtime win across canonical workloads.
- Production-scale Postgres benchmark runs completed on 2026-04-21 via one-shot canonical benchmark (`-benchmark-aggregations`, `driver=postgres`, with `runs_per_mode=1` and `runs_per_mode=3`).
- Run A (pre-tuning), daily window `2026-04-20T00:00:00Z` to `2026-04-21T00:00:00Z`: Go `4.000602432s` (`14881` rows) vs SQL `1h17m19.039092561s` (`14920` rows), with Go ~`1159.59x` faster.
- Run A (pre-tuning), monthly window `2026-04-01T00:00:00Z` to `2026-05-01T00:00:00Z`: Go `3.529410947s` (`15871` rows) vs SQL `3.313037973s` (`15873` rows), near parity with SQL slightly faster (~`0.216s`, `6.1%`).
- Run B (after PostgreSQL tuning), daily window `2026-04-21T00:00:00Z` to `2026-04-22T00:00:00Z`: Go `2.277889486s` (`14831` rows) vs SQL `1m31.273491543s` (`14839` rows), with Go still ~`40.07x` faster.
- Run B (after PostgreSQL tuning), monthly window `2026-04-01T00:00:00Z` to `2026-05-01T00:00:00Z`: Go `3.947474215s` (`15871` rows) vs SQL `2.758716002s` (`15873` rows), with SQL ~`1.43x` faster.
- Run C (after PostgreSQL tuning, `runs=3`), daily window `2026-04-21T00:00:00Z` to `2026-04-22T00:00:00Z`: Go avg `2.261369712s` (min `2.169537168s`, median `2.191474445s`, max `2.423097524s`, rows `14831`) vs SQL avg `1m31.738727387s` (min `1m29.960115863s`, median `1m32.068576507s`, max `1m33.187489791s`, rows `14839`), with Go ~`40.57x` faster by average.
- Run C (after PostgreSQL tuning, `runs=3`), monthly window `2026-04-01T00:00:00Z` to `2026-05-01T00:00:00Z`: Go avg `3.705308832s` (min `3.696553751s`, median `3.70776704s`, max `3.711605706s`, rows `15871`) vs SQL avg `3.065612298s` (min `2.873749798s`, median `3.022090149s`, max `3.300996948s`, rows `15873`), with SQL ~`1.21x` faster by average (~`17.26%` faster than Go).
- Tuning impact between Run A and Run B: daily SQL improved ~`50.83x`, daily Go improved ~`1.76x`, monthly SQL improved ~`1.20x`, and monthly Go regressed (~`0.89x` of prior speed).
- Decision remains unchanged: keep Go as scheduled default and treat SQL as fallback/backfill until SQL shows a clear, repeatable runtime win across canonical workloads, especially on daily windows (where Go remains consistently dominant across runs).
- [x] Keep Go as scheduled default unless SQL shows clear and repeatable runtime wins.
- [x] If SQL wins, roll out behind a controlled flag before any default switch.