Re: How is HBase used as a storage?

On Saturday, July 6, 2013 3:46:12 AM UTC+5:30, Ricky Saltzer wrote:

Hi Nancy -

While HBase uses HDFS to store a table's HFile, it's still a different
storage manager as far as Impala is concerned. HBase provides a different
set of APIs to get at data, for example, you can get a single record back
very quickly in HBase.

HBase also supports a notion range scans, Impala can take advantage of
this to minimize the amount of data it has to analyze. Impala makes use of
HBase's APIs to get at the data you're asking for, allowing you to analyze
data within HBase using SQL. The HBase storage engine also allows for
updates and deletes, something not possible with regular HDFS files. Impala
will later have the capability to perform updates/deletes/inserts directly
to HBase.

If you have data which is added and modified rapidly, you should go with
HBase. If you have data which doesn't change often, but is added
frequently, HDFS might be a better choice.

Here's some more about HBase: http://hbase.apache.org/

I hope this clears things up.



On Fri, Jul 5, 2013 at 4:13 AM, nancy jean <nancyj...@gmail.com<javascript:>


--
Ricky Saltzer
Tools Developer
http://www.cloudera.com

On Sun, Jul 7, 2013 at 10:12 PM, nancy jean wrote:

So, If I use just HBase as storage in Impala, where are all the data
getting stored? In which part of HBase is it getting stored? Or is it like
HBase will need HDFS to store the data?