You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@trafodion.apache.org by "Qifan Chen (JIRA)" <ji...@apache.org> on 2015/07/16 01:49:04 UTC

[jira] [Commented] (TRAFODION-41) Update statistics utility is too conservative in allowing in-memory sort operating mode

    [ https://issues.apache.org/jira/browse/TRAFODION-41?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14628954#comment-14628954 ] 

Qifan Chen commented on TRAFODION-41:
-------------------------------------

The solution includes the following steps. 

1. Perform a fit-test to determine if all sampled single and multiple rows (or internally represented as single and multi-column groups) can fit in memory. 
2. If the answer is positive, force the multi-column stats to be built in memory (currently it is controlled by the CQD USTAT_USE_INTERNAL_SORT_FOR_MC).
3. let the utility to build all the stats in memory. 
4. The hardcoded NAHeap estimated max size (1.3GB) is removed and a new CQD USTAT_NAHEAP_ESTIMATED_MAX is added to allow some flexibility to specify the amount of memory to use for internal sort. This memory is limited by the available physical memory to the process (tdm_arkcmp) in which the utility executes.  

> Update statistics utility is too conservative in allowing in-memory sort operating mode
> ---------------------------------------------------------------------------------------
>
>                 Key: TRAFODION-41
>                 URL: https://issues.apache.org/jira/browse/TRAFODION-41
>             Project: Apache Trafodion
>          Issue Type: Bug
>            Reporter: Qifan Chen
>            Assignee: Qifan Chen
>
> The update statistics is very conservative in use of  in-memory sort to create stats.  Currently, in memory mode is enabled only if no multi-column stats is requested. Under the default operation, 820MB memory is available for in-memory sort and can be more than adequate to build the multi-column stats in memory. 
> As a consequence, the utility will use a sample table to hold the sampled rows., which can be avoided with in-memory sort. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)