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 2016/01/28 23:10:40 UTC

[jira] [Assigned] (OAK-3916) Cor/Cow should copy in suggest dir as well if available

     [ https://issues.apache.org/jira/browse/OAK-3916?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Vikas Saurabh reassigned OAK-3916:
----------------------------------

    Assignee: Chetan Mehrotra  (was: Vikas Saurabh)

Talked to Chetan offline and it seems that we might not want to rush into this feature. So, moving this one to 1.6.

Along with that, he suggested that we might want to expose an index def level property - say {{useRamDirectory=true}} - to indicate that during IndexNode.open we might want to copy suggest data into a ram directory. [~chetanm], I was looking into ram directory doc and it seems to suggest a couple of issues - it doesn't perform well in multi threaded cases, (of course) uses up heap, might be unfruitful with several hundreds of MB. As an alternative, it suggests to use MMapDirectory - which afaics essentially boils down this issue itself (to have local FS dir and then backed by MMapDirectory).

So, I'm not sure if we should go for a ram directory option - I mean I think we still don't have enough data to take that call. Btw, I'm assigning the issue to you ([~chetanm]). Ping me if you feel that ram directory would be a useful addition.

> Cor/Cow should copy in suggest dir as well if available
> -------------------------------------------------------
>
>                 Key: OAK-3916
>                 URL: https://issues.apache.org/jira/browse/OAK-3916
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>            Reporter: Vikas Saurabh
>            Assignee: Chetan Mehrotra
>            Priority: Minor
>             Fix For: 1.4
>
>
> Suggest dictionary is now stored as another OakDir under lucene indices. That means that looking up for suggestions peeks into a lucene index which can get expensive if it requires going to NFS/Blobstore. It'd be nice to utilize Cor/Cow for this as well.



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