hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ChiaPing Tsai (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-15806) An endpoint-based export tool
Date Mon, 23 May 2016 19:45:13 GMT

    [ https://issues.apache.org/jira/browse/HBASE-15806?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15296957#comment-15296957

ChiaPing Tsai commented on HBASE-15806:

I don't profile all the phase but I measure the total job elapsed time (as [attached|https://issues.apache.org/jira/secure/attachment/12805500/Experiment.png]).
The endpoint approach and MR approach have many similarities as following:
1) same arguments (scan and output options) 
2) same split(one region one split)
3) same output (sequence file)
The factor that impacts the performance is the endpoint approach doesn't transfer data from
regionserver to client (save rpc), so the endpoint apprache will be faster than MR approach
(if the data output is not bottleneck)


> An endpoint-based export tool
> -----------------------------
>                 Key: HBASE-15806
>                 URL: https://issues.apache.org/jira/browse/HBASE-15806
>             Project: HBase
>          Issue Type: New Feature
>    Affects Versions: 2.0.0
>            Reporter: ChiaPing Tsai
>            Assignee: ChiaPing Tsai
>            Priority: Minor
>             Fix For: 2.0.0
>         Attachments: Experiment.png, HBASE-15806.patch
> The time for exporting table can be reduced, if we use the endpoint technique to export
the hdfs files by the region server rather than by hbase client.
> In my experiments, the elapsed time of endpoint-based export can be less than half of
current export tool (enable the hdfs compression)
> But the shortcomings is we need to alter table for deploying the endpoint
> any comments about this? thanks

This message was sent by Atlassian JIRA

View raw message