[jira] [Resolved] (IMPALA-2167) Remove the old (unpartitioned) HJ and AGG nodes

["Tim Armstrong (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/IMPALA-2167?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Tim Armstrong resolved IMPALA-2167.
-----------------------------------
       Resolution: Fixed
    Fix Version/s: Impala 2.10.0



IMPALA-4674: Part 1: remove old aggs and joins

This is intended to be merged at the same time as Part 2 but is
separated out to make the change more reviewable. Part 2 assumes
that it does not need special logic to handle this mode (e.g.
because the old aggs and joins don't use reservation).

Disable the --enable_partitioned_{aggregation,hash_join} options
and remove all product and test code associated with them.

Change-Id: I5ce2236d37c0ced188a4a81f7e00d4b8ac98e7e9
Reviewed-on: http://gerrit.cloudera.org:8080/7102
Reviewed-by: Tim Armstrong <ta...@cloudera.com>
Tested-by: Impala Public Jenkins

> Remove the old (unpartitioned) HJ and AGG nodes
> -----------------------------------------------
>
>                 Key: IMPALA-2167
>                 URL: https://issues.apache.org/jira/browse/IMPALA-2167
>             Project: IMPALA
>          Issue Type: Improvement
>          Components: Backend
>    Affects Versions: Impala 2.2
>            Reporter: Ippokratis Pandis
>            Assignee: Tim Armstrong
>            Priority: Minor
>             Fix For: Impala 2.10.0
>
>
> Currently we maintain two version of the hash-based aggregations and joins, the (old) unpartitioned ones and the partitioned and spillable ones. The main reason we had to keep the old version it was because of the additional memory PAGG and PHJ were consuming in small-ish aggregations and joins.
> But maintaining this extra code is cumbersome, error-prone and tricky to test. For example, the new PHJ supports functionality (join modes) that the old one does not support, which means that some times even though we disable PHJ we still use it, see IMPALA-1751.
> If we manage to make PAGG and PHJ to consume as much memory as their unpartitioned counterparts in small-ish inputs (or a few MBs more) then there is no reason we should keep the old AGG and HJ nodes around.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)