You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hbase.apache.org by "Bryan Duxbury (JIRA)" <ji...@apache.org> on 2008/03/13 17:48:25 UTC
[jira] Created: (HBASE-513) Store Mapfile indices in soft
references
Store Mapfile indices in soft references
----------------------------------------
Key: HBASE-513
URL: https://issues.apache.org/jira/browse/HBASE-513
Project: Hadoop HBase
Issue Type: Sub-task
Reporter: Bryan Duxbury
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
[jira] Updated: (HBASE-513) Store Mapfile indexes in soft
references
Posted by "Bryan Duxbury (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/HBASE-513?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Bryan Duxbury updated HBASE-513:
--------------------------------
Description:
One of the potentially big consumers of memory in the region server is the mapfile indexes we keep cached for fast access. Rather than building a complex system to try and LRU them ourself, we should use some form of SoftReference or derivative to keep the index around. That way, when our server gets near its memory limit, those indexes will be cleaned up automatically, making room.
We would also have to code our accesses of Mapfile in such a way that if a index had been garbage collected, we were able to fetch it again from disk with a minimum of trouble.
Summary: Store Mapfile indexes in soft references (was: Store Mapfile indices in soft references)
> Store Mapfile indexes in soft references
> ----------------------------------------
>
> Key: HBASE-513
> URL: https://issues.apache.org/jira/browse/HBASE-513
> Project: Hadoop HBase
> Issue Type: Sub-task
> Reporter: Bryan Duxbury
>
> One of the potentially big consumers of memory in the region server is the mapfile indexes we keep cached for fast access. Rather than building a complex system to try and LRU them ourself, we should use some form of SoftReference or derivative to keep the index around. That way, when our server gets near its memory limit, those indexes will be cleaned up automatically, making room.
> We would also have to code our accesses of Mapfile in such a way that if a index had been garbage collected, we were able to fetch it again from disk with a minimum of trouble.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.