Ignite SQL join performance

[Ray]

Currently the only way to join data is nested loops approach stated in this
two tickets.
https://issues.apache.org/jira/browse/IGNITE-6201
https://issues.apache.org/jira/browse/IGNITE-6202

So we can anticipate the SQL join performance will degrade a lot after the
data size reached certain amount.
After some research, I found out that deep down Ignite utilize H2 database
to do the join.
But currently H2 only supports nested loop approach to do the join.

Does it mean we'll have to remove H2 if we want to implement ticket 6201 and
6202?




--
Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/

[Vladimir Ozerov]

Hi Ray,

Basically, there are two ways:
1) Implement missing feature in H2, submit a patch, wait for new H2 version
to release, and then move Ignite to this version
2) Wait for our own execution engine to appear and then implement these
join algorithms there.

On Mon, Oct 30, 2017 at 9:49 AM, Ray <[hidden email]> wrote:

> Currently the only way to join data is nested loops approach stated in this
> two tickets.
> https://issues.apache.org/jira/browse/IGNITE-6201
> https://issues.apache.org/jira/browse/IGNITE-6202
>
> So we can anticipate the SQL join performance will degrade a lot after the
> data size reached certain amount.
> After some research, I found out that deep down Ignite utilize H2 database
> to do the join.
> But currently H2 only supports nested loop approach to do the join.
>
> Does it mean we'll have to remove H2 if we want to implement ticket 6201
> and
> 6202?
>
>
>
>
> --
> Sent from: http://apache-ignite-developers.2346864.n4.nabble.com/
>