[
https://issues.apache.org/jira/browse/HBASE-18752?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16196353#comment-16196353
]
Chia-Ping Tsai commented on HBASE-18752:
----------------------------------------
bq. Still would be nice to run PE random write tests for a bit longer duration
of say 10 mns to see the impact of the overhead in flush.
Run the test which creates 100 GB hfiles with following config
# snappy
# no compaction
# no wal
# 10 times
# DefaultMemStore
|| ||master||patch||
|min(s)|7255|7045|
|avg(s)|7491|7552|
|max(s)|7950|8030|
It seems the impact of the overhead is trivial.
> Recalculate the TimeRange in flushing snapshot to store file
> ------------------------------------------------------------
>
> Key: HBASE-18752
> URL: https://issues.apache.org/jira/browse/HBASE-18752
> Project: HBase
> Issue Type: Sub-task
> Reporter: Chia-Ping Tsai
> Assignee: Chia-Ping Tsai
> Fix For: 2.0.0-beta-1
>
> Attachments: HBASE-18752.v0.patch, HBASE-18752.v1.patch
>
>
> We drop superfluous cells in flushing, hence the TimeRange from snapshot is
> inaccurate for the storefile. We should recalculate the TimeRange for the
> storefile, but the side-effect is the extra cost - we need to extract the
> timestamp from cell (ByteBufferCell).
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
