[
https://issues.apache.org/jira/browse/HBASE-18394?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16129501#comment-16129501
]
Ethan Wang commented on HBASE-18394:
------------------------------------
[~ted_yu]For example, for storing naive time series data, my row key looks like:
{code:java}
Scope.Metric.Timestamp
{code}
In this case, there are "three levels" of prefix
Scope
Metric
TimeStamp.
Rather than
S, Sc, Sco, Scop, ....
> Allowing user to define levels of prefix of Rowkey to be also put into
> BloomFilter
> ----------------------------------------------------------------------------------
>
> Key: HBASE-18394
> URL: https://issues.apache.org/jira/browse/HBASE-18394
> Project: HBase
> Issue Type: Improvement
> Reporter: Ethan Wang
>
> For Time Series data, one common use case is get/scan by prefix of a row key.
> Such as:
> RowKey: Scope.Metric.TimeStamp
> Search by: Scope*
> In these case bloom filter is not leveraged. However if we store every prefix
> of row key into bloomfilter that will be a big over head. Let user to store
> only the prefix of certain level will be a good thing to have
>
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
