Learned Indexes in PostgreSQL?

Started by Stefan Kellerabout 8 years ago3 messages
#1Stefan Keller
sfkeller@gmail.com

Hi,

This is an awesome paper about a new index called "Learned Index".
it's aka dense hash structure derived ("learned") from actual data.
Looks very promising for certain properties [*].

Anyone already working on this in Postgres?
How could this be implemented in Postgres?

:Stefan

[*] "The Case for Learned Index Structures" Kraska et al. (Dec 2017)
https://arxiv.org/abs/1712.01208

#2Robert Haas
robertmhaas@gmail.com
In reply to: Stefan Keller (#1)
Re: Learned Indexes in PostgreSQL?

On Tue, Dec 12, 2017 at 6:38 AM, Stefan Keller <sfkeller@gmail.com> wrote:

This is an awesome paper about a new index called "Learned Index".
it's aka dense hash structure derived ("learned") from actual data.
Looks very promising for certain properties [*].

Anyone already working on this in Postgres?
How could this be implemented in Postgres?

You're the third person to start a thread on this topic in the last few days.

Here are links to the other two threads:

/messages/by-id/CANNMO+J1KeTSx5q5SYuwHf1v-gPRLrOZw1s7qOpqWx=3UMMvtg@mail.gmail.com
/messages/by-id/CAAerrx8bHiW3rgAqpoLqjhYhk7gHOrDtkE5UC9DAsv1w5FzpEw@mail.gmail.com

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

#3Stefan Keller
sfkeller@gmail.com
In reply to: Robert Haas (#2)
Re: Learned Indexes in PostgreSQL?

2017-12-12 21:26 GMT+01:00 Robert Haas <robertmhaas@gmail.com>:

Here are links to the other two threads:

/messages/by-id/CANNMO+J1KeTSx5q5SYuwHf1v-gPRLrOZw1s7qOpqWx=3UMMvtg@mail.gmail.com
/messages/by-id/CAAerrx8bHiW3rgAqpoLqjhYhk7gHOrDtkE5UC9DAsv1w5FzpEw@mail.gmail.com

Thanks, Robert.
My fault not having searched before this list - in fact I did but inaccurately.

:Stefan

2017-12-12 21:26 GMT+01:00 Robert Haas <robertmhaas@gmail.com>:

Show quoted text

On Tue, Dec 12, 2017 at 6:38 AM, Stefan Keller <sfkeller@gmail.com> wrote:

This is an awesome paper about a new index called "Learned Index".
it's aka dense hash structure derived ("learned") from actual data.
Looks very promising for certain properties [*].

Anyone already working on this in Postgres?
How could this be implemented in Postgres?

You're the third person to start a thread on this topic in the last few days.

Here are links to the other two threads:

/messages/by-id/CANNMO+J1KeTSx5q5SYuwHf1v-gPRLrOZw1s7qOpqWx=3UMMvtg@mail.gmail.com
/messages/by-id/CAAerrx8bHiW3rgAqpoLqjhYhk7gHOrDtkE5UC9DAsv1w5FzpEw@mail.gmail.com

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company