Here in Uber, we are working on a project on sampling Kafka stream data
and user query log to decide whether to index columns and other table
config like dictionary encoding. This is very useful for self on-boarding
Pinot tables. The main focus now is to add index. But if we can remove
index too, there will be helpful too.

On Tue, Feb 11, 2020 at 5:12 PM kishore g <g.kish...@gmail.com> wrote:

> It's not trivial for users to know when to add an index. It definitely
> improves the experience if they can easily add/remove indexes. Regarding
> storage, indexes can be quite big depending on the cardinality for e.g. inv
> index on something like memberId is not very useful.
>
> On Tue, Feb 11, 2020 at 5:07 PM Mayank Shrivastava <
> mayanks.apa...@gmail.com>
> wrote:
>
> > Subbu, I think the proposal is to reduce the storage cost, which is why I
> > asked how often is that seen (outside LinkedIn).
> >
> > Sent from my iPhone
> >
> > > On Feb 11, 2020, at 3:38 PM, Subbu Subramaniam <mcvsu...@apache.org>
> > wrote:
> > >
> > > I believe we do not map the index if it is not sset in the table
> > config, so the index is not loaded.
> > >
> > > The cost on storage exists, but hopefully is a lot smaller than memory.
> > >
> > > -Subbu
> > >
> > >> On 2020/02/11 22:37:13, kishore g <g.kish...@gmail.com> wrote:
> > >> Currently, we only support adding indices to the pinot segment. If the
> > >> index is removed from table config, we dont load it but continue to
> > keep it
> > >> in the segment. The only drawback with this is additional usage of
> disk
> > >> space.
> > >>
> > >> We should enhance the segment preprocessor to have the ability to
> > rewrite
> > >> the segment.
> > >>
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: dev-unsubscr...@pinot.apache.org
> > > For additional commands, e-mail: dev-h...@pinot.apache.org
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscr...@pinot.apache.org
> > For additional commands, e-mail: dev-h...@pinot.apache.org
> >
> >
>

Reply via email to