[
https://issues.apache.org/jira/browse/HBASE-15554?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15215441#comment-15215441
]
Anoop Sam John commented on HBASE-15554:
----------------------------------------
You mean this one
{code}
bloomKeyKV = KeyValueUtil.createFirstOnRow(cell.getRowArray(),
cell.getRowOffset(),
cell.getRowLength(),
HConstants.EMPTY_BYTE_ARRAY, 0, 0, cell.getQualifierArray(),
cell.getQualifierOffset(),
cell.getQualifierLength());
{code}
We need to create a new byte[] with rk and Q alone and that is needed.. Can
avoid the create of KV object every time.. But there is a check for increasing
KV compared to previous written cell.. Can we avoid this check altogether
now? If so, we can make some change in the bloom write f/w so as to pass a
Cell directly and the writer decides what all part of that cell to be
considered for the bloom write. (?)
> 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: Vladimir Rodionov
> Fix For: 2.0.0
>
>
> Accounts for 10% memory allocation when BloomFilterType is ROWCOL.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
