[
https://issues.apache.org/jira/browse/HBASE-7845?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13609849#comment-13609849
]
Liang Xie commented on HBASE-7845:
----------------------------------
Hi [~stack], the expected blocks read number being reduced because:
Without patch, the search for row/cf1/col5 is done by looking for
"row/cf1/col5/LATEST_TIMESTAMP/...", and this will land us in previous
block(col4), found mis-matched(blockseek return 1), then advance to current
block(issued a HFileScanner.next operation).
With patch, due to faked index key effect, we locate the current block(col5)
directly and blockSeek return -2, then we don't need to request
HFileScanner.next, which will call readNextDataBlock internally
Hope it's clear, sorry for my poor english:)
> optimize hfile index key
> ------------------------
>
> Key: HBASE-7845
> URL: https://issues.apache.org/jira/browse/HBASE-7845
> Project: HBase
> Issue Type: Improvement
> Components: HFile
> Affects Versions: 0.96.0
> Reporter: Liang Xie
> Assignee: Liang Xie
> Fix For: 0.98.0
>
> Attachments: HBASE-7845.txt, HBASE-7845-v2.txt, HBASE-7845-v3.txt,
> HBASE-7845-v4.txt, HBASE-7845-v5.txt, HBASE-7845-v6.txt, HBASE-7845-v7.txt,
> HBASE-7845-v8.txt
>
>
> Leveldb uses ByteWiseComparatorImpl::FindShortestSeparator() &
> FindShortSuccessor() to reduce index key size, it would be helpful under
> special conditions.
--
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
