[
https://issues.apache.org/jira/browse/BLUR-5?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Patrick Hunt updated BLUR-5:
----------------------------
Attachment: BLUR-5.patch
My initial patch for supporting the write through cache. This is against the
lucene-4.0.0 branch. It's basically functional, I've run the unit tests
successfully but not yet had a chance to run the real server against this code.
Would appreciate any review/comments.
Notice: currently everything is write cached, similar to read
testing:
1) added a test for the write cache by using a larger cache, this results in
everything written (and subseq read) being in the cache. fwiw I've seen perf
improve by ~30-40% on my machine when using the same seed.
2) I had to remove the "setLength" calls in the createFile method - it's not
guaranteed that those methods are implemented, and when they are there is no
guarantee what is actually in the file if not explicitly written. In some cases
the test would write less than it had set. This works ok with the prior
implementation which is calling file.setLength (although it's not clear why
it's successful given the file.setLength says the content is undefined if not
specifically written)
> Write through caching for the BlockCache
> ----------------------------------------
>
> Key: BLUR-5
> URL: https://issues.apache.org/jira/browse/BLUR-5
> Project: Apache Blur
> Issue Type: Improvement
> Reporter: Aaron McCurry
> Attachments: BLUR-5.patch
>
>
> This will allow for better NRT update performance because the writer will not
> have to read the NRT segments from HDFS.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
