crunch-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Micah Whitacre (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CRUNCH-230) Attempt to estimate HBase table sizes when we're given a trivial Scan object
Date Tue, 12 Nov 2013 03:20:20 GMT

    [ https://issues.apache.org/jira/browse/CRUNCH-230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13819780#comment-13819780
] 

Micah Whitacre commented on CRUNCH-230:
---------------------------------------

The patch looks fine.  I wonder if we can handle additional cases to get better size estimates.
 For example you check that the table is scanned from the beginning to the end but not the
entire width.  The thought I had is you could get check for only the column families associated
with the scan.  There isn't a nice API to expose those paths but they typically are /hbase/<table_name>/<region>/<column
family>[1]  So we could pass in a regex ignoring the region value to get estimates per
column family.

[1] - http://hbase.apache.org/book/trouble.namenode.html

> Attempt to estimate HBase table sizes when we're given a trivial Scan object
> ----------------------------------------------------------------------------
>
>                 Key: CRUNCH-230
>                 URL: https://issues.apache.org/jira/browse/CRUNCH-230
>             Project: Crunch
>          Issue Type: Improvement
>          Components: IO
>    Affects Versions: 0.6.0
>            Reporter: Josh Wills
>            Priority: Minor
>         Attachments: CRUNCH-230.patch
>
>
> If we're asked to do a scan of an entire HBase table, we can actually do a pretty good
job of estimating how large it is by looking up its directory in HDFS. This patch checks the
input scan given to the HBaseSourceTarget, and if it doesn't specify any filters, looks up
the size of the input table on HDFS.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Mime
View raw message