4 ms·
>They did not graph write speed in the article though, which was a bit disappointing, so I assume the difference was negligible Did you mean query performance?
by RobAtticus 8y ago
>They did not graph write speed in the article though, which was a bit disappointing, so I assume the difference was negligible
Did you mean query performance? There are write speed/insert numbers under "Insert performance measurement graphs"
- hardwaresofton 8y agoOh sorry, I somehow focused in on the number of partitions bit and somehow missed that the metric was actually insert performance. I'm surprised PG10 tanked so fast...
- anarazel 8y agoThe partition selection / pruning in 10 was not yet improved from older inheritance based "partitioning" (COPY might already have been optimized back then, not sure right now). It's looking different in v11.
- hardwaresofton 8y agoAhhh I thought the partitioning was NOT inheritance based by default after reading the partition documentation page in the PG docs[0]. More specifically the "Implementation using inheritance" section[1] -- I thought the default was NOT to use table inheritance, but that you could use it as an option. [0]: https://www.postgresql.org/docs/current/static/ddl-partitioning.html https://www.postgresql.org/docs/current/static/ddl-partition... [1]: https://www.postgresql.org/docs/current/static/ddl-partitioning.html#DDL-PARTITIONING-IMPLEMENTATION-INHERITANCE https://www.postgresql.org/docs/current/static/ddl-partition...
- anarazel 8y agoYou misunderstood me, sorry I should have been clearer. Some internals are shared between inheritance and declarative partitioning. For v10 most parts of the system (except for COPY) used the same partition exclusion logic for both forms of partitioning. Which means it was O(partitions) with a pretty high constant.