You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ignite.apache.org by "Vladimir Ozerov (JIRA)" <ji...@apache.org> on 2017/08/16 06:52:00 UTC
[jira] [Created] (IGNITE-6072) SQL: client node should not hold the
whole data set in-memory when possible when merge table is there
Vladimir Ozerov created IGNITE-6072:
---------------------------------------
Summary: SQL: client node should not hold the whole data set in-memory when possible when merge table is there
Key: IGNITE-6072
URL: https://issues.apache.org/jira/browse/IGNITE-6072
Project: Ignite
Issue Type: Task
Components: sql
Affects Versions: 2.1
Reporter: Vladimir Ozerov
Assignee: Alexander Paschenko
Fix For: 2.2
We implemented a fix for a case without merge table (IGNITE-6019). Let's try to do the same for merge table case. Essentially, we need to split all queries into two groups:
- Blocking - i.e. it require the whole result set from mappers first
- Non-blocking - everything what is not blocking :-)
Fix should be applied only to non-blocking case. Raw estimation from my side, what is "blocking":
1) GROUP BY/HAVING
2) DISTINCT
3) LIMIT/OFFSET
4) Subqueries
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)