[
https://issues.apache.org/jira/browse/GORA-116?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13251497#comment-13251497
]
Eric Newton commented on GORA-116:
----------------------------------
If I know in advance the distribution of my table, I can split it early to
increase the speed at which I can ingest new data. I just used the shortest
representation of the splits I wanted.
Accumulo uses the minimum difference between two rows as its natural split
point in order to keep the metadata about splits small. Keith worked around
this in the Accumulo back-end.
It's a little more complicated than just padding. If I split at \x10, \x10\x00
comes after \x10, which puts it in the next tablet. It will probably all just
work since we're just providing a locality hint, and a little spillage off the
ends of the tablet isn't going to impact performance that much.
> gora treats split points as if they represent actual values in the table
> ------------------------------------------------------------------------
>
> Key: GORA-116
> URL: https://issues.apache.org/jira/browse/GORA-116
> Project: Apache Gora
> Issue Type: Bug
> Components: storage-hbase
> Reporter: Eric Newton
> Priority: Minor
>
> Doing goraci testing with the hbase back-end for gora. I created single-byte
> split points. When I tried to map-reduce over the table, the gora back-end
> failed trying to convert the split points into Longs.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
