Changes between Initial Version and Version 1 of Ticket #31202, comment 11
- Timestamp:
- Feb 28, 2022, 6:16:10 AM (3 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #31202, comment 11
initial v1 8 8 - `bulk_update` shows much worse runtime behavior than all others (thus accounting is stopped early) 9 9 - `django-bulk-update` can keep up a bit longer, but then shows the same worse runtime growth (prolly due to the same SQL logic used?). This gets really worse for updates >5k (not shown). 10 - `copy_update` has a much higher setup costs (1 to 256 updated records)10 - `copy_update` has much higher setup costs (1 to 256 updated records, kinda expected due to temp table and index creation) 11 11 - between 4096 to 8192 updates `copy_update` starts to pay off compared to `fast_update`, at 32k updates it is almost twice as fast 12 - not shown in the charts: for higher update scounts it keeps gaining ground (almost being 4 times faster for 1M update records)12 - not shown in the charts: for higher update counts it keeps gaining ground (almost being 4 times faster for 1M update records) 13 13 - There is something going on between 256-512 updates, as almost all implementations show a steep jump there (postgres b-tree depth change? did not investigate that further...) 14 14