github-actions[bot]
commented
2 years ago Benchmark results
| Part | Duration |
|---|---|
| Create 5000 new messages | 385.02ms |
| Validate 5000 messages | 476.66ms |
| Native to db format 5000 messages | 109.92ms |
| Db to native format 5000 messages | 81.47ms |
| Add 1000 elements | 495.01ms |
| Add 1000 box1 msgs | 1002.72ms |
| Unbox 1000 box1 msgs first run | 164.14ms |
| Unbox 1000 box1 msgs second run | 106.55ms |
| Add 1000 box1 msgs | 967.17ms |
| Query 1000 msgs first run | 38.19ms |
| Query 1000 msgs second run | 21.54ms |
| Add 1000 box2 msgs | 1288.51ms |
| Unbox 1000 box2 msgs first run | 184.99ms |
| Unbox 1000 box2 msgs second run | 153.66ms |
| Migrate (+db1) | 10838.25ms |
| Migrate (alone) | 3717.58ms |
| Migrate (+db1 +db2) | 8095.41ms |
| Migrate (+db2) | 5646.02ms |
| Migrate continuation (+db2) | 871.86ms |
| Memory usage without indexes | 741.70 MB = 35.31 MB + etc |
| Initial indexing | 606.77ms |
| Initial indexing maxcpu=86 | 3831.82ms |
| Initial indexing compat | 630.93ms |
| Two indexes updating concurrently | 936.58ms |
| Key one initial | 46.98ms |
| Key two | 0.49ms |
| Key one again | 0.90ms |
| Reboot and key one again | 51.07ms |
| Latest root posts | 739.77ms |
| Latest posts | 8.81ms |
| Votes one initial | 512.04ms |
| Votes again | 0.48ms |
| HasRoot | 355.56ms |
| HasRoot again | 0.32ms |
| Author one posts | 462.54ms |
| Author two posts | 20.45ms |
| Dedicated author one posts | 493.06ms |
| Dedicated author one posts again | 0.48ms |
| DeleteFeed | 2575.35ms |
| Maximum memory usage | 1017.72 MB = 64.14 MB + etc |
| Indexes folder size | 10.01mb |
Fixes (hopefully) issue #380.
I had difficulty reproducing this bug in tests. We would need AAOL compactionProgress to emit more often than 1sec (currently it's hard coded) and we would need to have a way to immediately kill the sbot (to simulate a crash) but we can't because
sbot.closewill calllog.closewhich actually waits for compaction to complete.