[
https://issues.apache.org/jira/browse/HBASE-15554?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
ramkrishna.s.vasudevan updated HBASE-15554:
-------------------------------------------
Attachment: HBASE-15554_3.patch
An much more advanced version of the patch.
-> Creates an HasKey interface and this represents cells which will have key
part in contiguous structure - be it onheap or offheap.
-> Currently KeyValue, OffheapKeyValue, DBE cells (except PRefix Tree) fall
under this category.
-> Patch helps to avoid garbage for both ROW and ROW_COL cases.
-> Important thing to be noted is that the Bloom key will have the family part
added to it which was not there previously. Without this we cannot avoid the
copy in this Bloom write path.
-> Hash algos work with BBs also now.
-> Some code duplication seems to be unavoidable and refactoring was very
tricky considering the fact that the lastCell, lastBloomKey had to be tracked
for adding to the fileinfo.
Tested with Offheap BC, Offheap with DBE, onheap DBE etc. Will do some more
testing.
> 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, HBASE-15554_3.patch
>
>
> Accounts for 10% memory allocation in compaction thread when BloomFilterType
> is ROWCOL.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
