[
https://issues.apache.org/jira/browse/HBASE-15554?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15277633#comment-15277633
]
ramkrishna.s.vasudevan commented on HBASE-15554:
------------------------------------------------
For ROWCOL things unless we change things to work with Cell even in the hash
algo we cannot avoid that copy.
I will come up with an initial patch to work with cells but few things to keep
in mind, once we move to cell then we cannot assume everything is in a single
byte[] or BB. So how do we make the hash algo to work? Currently since the key
byte[] is getting used the byte representing the keylength, rowlength,
familylength is also getting used in the hash algo. On changing it to cell can
we still go with that assumption? First let me make a patch with Cell and see
how it works out.
> StoreFile$Writer.appendGeneralBloomFilter generates extra KV
> ------------------------------------------------------------
>
> Key: HBASE-15554
> URL: https://issues.apache.org/jira/browse/HBASE-15554
> Project: HBase
> Issue Type: Sub-task
> Components: Performance
> Reporter: Vladimir Rodionov
> Assignee: ramkrishna.s.vasudevan
> Fix For: 2.0.0
>
> Attachments: HBASE-15554.patch
>
>
> Accounts for 10% memory allocation in compaction thread when BloomFilterType
> is ROWCOL.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
