[
https://issues.apache.org/jira/browse/ORC-997?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Dongjoon Hyun updated ORC-997:
------------------------------
Affects Version/s: 1.8.0
> Share compress output buffer among TreeWriter
> ---------------------------------------------
>
> Key: ORC-997
> URL: https://issues.apache.org/jira/browse/ORC-997
> Project: ORC
> Issue Type: New Feature
> Components: compression, Java
> Affects Versions: 1.8.0
> Reporter: Lei Sun
> Assignee: Lei Sun
> Priority: Major
>
> There's individual instance of `OutStream` within each TreeWriter created by
> `WriterContext#createStream` method. Within `OutStream`, there are totally 3
> buffers:
> * current: the regular input buffer holding uncompressed, unencrypted bytes.
> * compress: the output buffer holding compressed bytes
> * overflow: same as "compress" but only used when the last compression
> output is larger than remaining capacity of compress buffer.
> Potentially the compress and overflow buffer doesn't have to be allocated
> individually within each OutStream object, but shared across all of them so
> to save memory. This ticket is trying to achieve that and obtain saving
> number with some sort of benchmarking(yet to be defined).
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
