OH: "PostgreSQL is the Linux of databases" It's probably equally unstoppable and will completely dominate the database ecosystem in the next 10-20 years. What's your opinion on this?
There's plenty more. Not sure about autovacuum in particular however: While our scheduling around it needs improvement, there's also some substantial gains from not needing full undo. FWIW, it's definitely fixable. There are actually quite a few patches around undo/zheap out.
-
-
Autovacuum actually had a big problem, don't know if it changed in the last 2-3 major versions: it wasn't triggered for INSERT only tables. By design, I think. But this was a problem for index visibility map (at least, this was my non-expert analysis of a performance problem)
-
Yea, but that's just because we're slow to improve, it's not an inherent issue.
- 2 more replies
New conversation -
-
-
It's more the need for vacuum, CoW, transactions wraparound, etc. I'd be happy if some day the way PostgreSQL writes data changes.
-
Sure, got that. Just think that the cost of other approaches are often underestimated. And I've spent a fair bit over the last ~2-3 years working around allowing different approaches to work in PG. Some of the costs of what you mentioned aren't costs intrinsic to in-heap mvcc.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
