[jira] [Resolved] (IMPALA-4698) Optimization for query like "select * from table limit 1"

["Alexander Behm (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/IMPALA-4698?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Alexander Behm resolved IMPALA-4698.
------------------------------------
    Resolution: Information Provided

EXEC_SINGLE_NODE_ROWS_THRESHOLD addresses this problem.

> Optimization for query like "select * from table limit 1"
> ---------------------------------------------------------
>
>                 Key: IMPALA-4698
>                 URL: https://issues.apache.org/jira/browse/IMPALA-4698
>             Project: IMPALA
>          Issue Type: Improvement
>          Components: Distributed Exec
>    Affects Versions: Impala 2.5.0
>            Reporter: Juan Yu
>
> Query like Select * from table limit 1 doesn't have any filter, the plan will include all partitions and plan fragment could be large. On large cluster and large table, plan distribution could take the majority of time. much longer than real query execution time.
> Possible optimization:
> 1.  For such query, planner could set num_nodes to a very small number automatically so the plan doesn't need to be sent to all hosts.
> 2. So far, query cannot be completed if plan fragment distribution is not done. with limit 1, as long as one plan fragment is complete, the query is done, no need to continue distribute plan fragments to rest of nodes. coordinator could check query progress before the distribute plan fragment RPC and simply mark the fragments are completed and return.



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