[teiid-issues] [JBoss JIRA] Commented: (TEIID-1750) Memory handling improvements
Steven Hawkins (JIRA)
jira-events at lists.jboss.org
Wed Sep 21 12:45:26 EDT 2011
[ https://issues.jboss.org/browse/TEIID-1750?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12629689#comment-12629689 ]
Steven Hawkins commented on TEIID-1750:
---------------------------------------
Changed the buffermanager to only be concerned with the memory state. Changed the buffering algorithm to be a tiered lru, which significantly reduces reads/writes when accessing table structures. the old lru/mru/working set model was specifically designed for tuplebuffers which have a linear insertion and forward iteration model. there will still need to be some tuning made to better support tuplebuffers.
tuplebuffers/tables no longer directly hold managedbatches, which reduces the overall memory footprint. This also helps effectively limit the softreference population.
put in a rough cut of a block storage scheme. this is likely to be further refined. the code should naturally allow for off-heap direct memory buffers to be used for large vm scenarios.
> Memory handling improvements
> ----------------------------
>
> Key: TEIID-1750
> URL: https://issues.jboss.org/browse/TEIID-1750
> Project: Teiid
> Issue Type: Quality Risk
> Components: Query Engine
> Affects Versions: 7.0
> Reporter: Steven Hawkins
> Assignee: Steven Hawkins
> Fix For: 7.6
>
>
> The buffermanager will create softreferences for batches that prefer memory, but does not track the memory as held once the value has been passivated.
> Also small lobs, such as xml values created via sqlxml, should be inlined into results for greater performance - and so that they can be paged out by the buffermanager rather than being held by a lobmanager.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
More information about the teiid-issues
mailing list