Why HBase backed Hive table uses MapReduce

Question

I am using Hbase backed Hive tables in my project but the reason we opted for Hbase backed Hive is to perform Updates.

Apart from that what are the other advantages of Hbase backed Hive tables. As it still uses MapReduce when queried from Hive. Even if we want small set of Data and as the table is Huge it takes time to give the result.

But if we perform a Scan with Range or Just a get in Hbase on Hbase shell results come in fraction of seconds. So what are the other advantages of using Hbase backed Hive table apart from updates(which is now available in HIVE as well) & SQL ease.

How does HIVE evaluates and Runs a Query if it is backed by Hbase ? Why it uses MapReduce to scan & give result instead of Hbase engine which is much faster ?? And does Hbase has its own engine to perform Scan, get operations to fetch data from its HFiles ???

Answer 1

I will advise you not to use Hbase backed Hive. As you can see the scan with filter runs in friction of the time that hive query runs. That because Hbase filter the data in the storage level and hive load all the table data and then filter it. There were suppose to be predicate pushdown from hive to Hbase, but there are lot of open issue in matter. And a lot of the predicate pushdown is disable. For more you can check the page : Hive HBase Integration