HNHacker News
TopNewBestAskShowJobs

akorotkov

188 karma · joined April 30, 2021

submissionscomments
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
Feel free to contact me at alexander@orioledb.com
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
Please, see this thread of comments. https://news.ycombinator.com/item?id=30466322
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
From my point of view, PostgreSQL development is quite conservative. Every small change may become a subject of heated debates, nothing to say about big changes. This approach to the development have advantages and disadvantages. But if we imagine OrioleDB accepted to the upstream, that would be most dramatical changes since PostgreSQL was released to Open Source. This is why I decided to make a fork and then merge it to upstream step-by-step.

It's not that bad to use a fork during 2-3 release cycles give that: 1) It's not highly divergent fork (the patchset is small). 2) They patchset will be smoothly decrease over that period of time. Also, even 2-3 release cycles isn't that bad in comparison with decades while we have small progress in wicked problems.

akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
PostgreSQL 32-bit transaction identifiers are cycled in a ring. This is why PostgreSQL needs to "freeze" old transaction identifiers before they could start being identified as future. This routine is "vacuum freeze" and it have to scan and re-write pages, which could have no dead rows.
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
Additionally OrioleDB fundamentally changes design of buffer management and improves the vertical scalability a lot. https://www.slideshare.net/AlexanderKorotkov/solving-postgre...

Also, I have to mention that zheap development is inactive for now. And its undo implementation was never complete. Zheap undo works only if you don't update any indexed column. Otherwise, it works as plain heap with write-amplification and bloat.

akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
> That has its own set of tradeoffs - if you're running a long transaction with higher level of serialization, the database has to look for old data in the undo segment which is not only likely vastly slower than normal table, but might not even have the data anymore (any ETL developer have probably seen "Snapshot too old" error from Oracle).

I would say "Snapshot too old" isn't essential drawback of undo log. DBMS can keep the the undo log records until all snapshots, which needs them, are released. And this is how OrioleDB behaves. This is also kind of "bloat", but cleanup is cheaper. You just have to cut unclaimed undo log instead of expensive vacuum.

akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
There is not exactly a documentation. But there you can see that commits after "Stamp 14.2." are extendability patches. Commit messages briefly describe the changes. https://github.com/orioledb/postgres/commits/patches14 Patches will be polished and published on pgsql-hackers.
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
Hi! My name is Alexander Korotkov, I'm founder of OrioleDB. Thank you for your feedback and support. Please, write me at alexander@orioledb.com Thank you!
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
I mean major releases. So, targeting 17.0.
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
Right. The idea of active-active OrioleDB multimaster is to apply changes locally and in parallel send it to the leader. Then sync on commit and ensure there is no conflicts.

The design document will be published later.

akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
Right, no more vacuums. And within typical workloads there should be no bloat.
akorotkov··on OrioleDB – solving some PostgreSQL wicked problems
Yes, that's it.

1. The first goal is to become a pure extension. That should be done in 2-3 PostgreSQL release cycles. 2. The long term goal is to become a part of PostgreSQL.

← PreviousPage 2 of 2