You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-issues@jackrabbit.apache.org by "Vikas Saurabh (JIRA)" <ji...@apache.org> on 2017/10/04 08:06:01 UTC
[jira] [Commented] (OAK-3132) Reduce memory usage of OakIndexFile
[ https://issues.apache.org/jira/browse/OAK-3132?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16190953#comment-16190953 ]
Vikas Saurabh commented on OAK-3132:
------------------------------------
I think with CopyOnRead and OAK-6269, this should no longer be an issue. [~mreutegg], [~chetanm], may be we should close this issue?
> Reduce memory usage of OakIndexFile
> -----------------------------------
>
> Key: OAK-3132
> URL: https://issues.apache.org/jira/browse/OAK-3132
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: lucene
> Reporter: Marcel Reutegger
>
> OakIndexFile splits an index file into Blob chunks and keeps one of the chunks in memory at a time. The chunk size is 32k by default but can be controlled with a property on the index definition.
> In a test repository I observed OakIndexFile instances consuming more than 600MB of heap, with a chunk size of 1MB. This seems quite high for a system where no query was active at that time.
> I think we should look into alternative ways to read from the Blob than just plain old InputStream. E.g. with a ByteChannel we could push the caching aspect to the Blob implementation, which could leverage an OS buffer cache.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)