[
https://issues.apache.org/jira/browse/IGNITE-13027?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17111099#comment-17111099
]
Andrey Mashenkov commented on IGNITE-13027:
-------------------------------------------
I'd remove all segment mentions from SQL protocol. Node doesn't need to know
about remote segments or specify segment for "map" query.
We introduced segmented index because it is the only solution to traverse index
in parallel without serious hacks in depth of H2.
This brought (pain) a big limitation to SQL JOINs and force users to have
"query parallelism" turn on for all caches that ever could be participate in
JOINs.
We should rework index segmentation in new SQL engine and remove 'segments'
from protocol.
We can keep segments as a part of storage implementation with making query flow
independent from segments or even implement some algorithm for traverse index
in parallel threads.
To support compatibility with segmented indices (if user already have ones), we
can make a wrapper for segmented index that will act as non-segmented index and
use kind of merge-sort underneath.
This will allow user to have different segmentation levels for different
indices and correct JOINs at same time.
> Calcite integration. Query parallelism support.
> -----------------------------------------------
>
> Key: IGNITE-13027
> URL: https://issues.apache.org/jira/browse/IGNITE-13027
> Project: Ignite
> Issue Type: Improvement
> Components: sql
> Reporter: Roman Kondakov
> Priority: Major
>
> We need to add support for segmented indexes in Calcite engine. Currently
> engine is able to execute queries over index with only one segment.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
