Обсуждение: AI for query-planning?
Hi, are there any plans for using some kind of AI for query-planning?
Can someone with more knowledge about this than I have please explain why it might, or not, be a good idea, and what the challenges are?
Thanks.
--
Andreas Joseph Krogh
CTO / Partner - Visena AS
Mobile: +47 909 56 963
Вложения
On 6/22/24 04:50, Andreas Joseph Krogh wrote: > Hi, are there any plans for using some kind of AI for query-planning? > > Can someone with more knowledge about this than I have please explain > why it might, or not, be a good idea, and what the challenges are? 1) Require large amount of resources. 2) Produce high rate of incorrect answers. > > Thanks. > > -- > *Andreas Joseph Krogh* > CTO / Partner - Visena AS > Mobile: +47 909 56 963 > andreas@visena.com <mailto:andreas@visena.com> > www.visena.com <https://www.visena.com> > <https://www.visena.com> -- Adrian Klaver adrian.klaver@aklaver.com
On Sat, Jun 22, 2024, 5:20 PM Andreas Joseph Krogh <andreas@visena.com> wrote:
Hi, are there any plans for using some kind of AI for query-planning?
Can someone with more knowledge about this than I have please explain why it might, or not, be a good idea, and what the challenges are?
not totally ai, but it can use data from your database to build some metadata.
there are also plan analyzers online which people paste their plan to get a quick summary of problem areas, that data can be useful too for ai based plan optimisers.
but that said, postgresql has a lot of knobs to tune, so ...
## Andreas Joseph Krogh (andreas@visena.com): > Hi, are there any plans for using some kind of AI for query-planning? Actually, we do have our GEQO - Genetic Query Optimization - already in the planner: https://www.postgresql.org/docs/current/geqo.html As per the common taxomonies, genetic algorithms are part of the larger "Artificial Intelligence" domain - they're just not part of the current hype cycle (some of us still remember the early iterations of neuronal networks, or fuzzy logic, or...). > Can someone with more knowledge about this than I have please explain why it > might, or not, be a good idea, and what the challenges are? Just waving your hand and saying "AI" three times in a row does not manifest an usable algorithm. Going the full way from a problem to be understood and solved to algorithm to implementation requires more effort - and especially with all the "hot off the hype press" ideas that already fails in the very first stages. Remember just recently, when some people were "predicting" that all databases would become blockchains? Regards, Christoph -- Spare Space