1. Integrations Guide
  2. Integration with Apache Hive
  3. Mapping the Hive Data Model to the Oracle NoSQL Database Table Model
  4. YARN Versus MapReduce Version 1

YARN Versus MapReduce Version 1

Hadoop deployments can include two versions of MapReduce. The first version (referred to as MRv1) is the original version of MapReduce; and consists of interfaces and classes from the Java package org.apache.hadoop.mapred. The newer version of MapReduce is referred to as YARN (Yet Another Resource Negotiator) or, more generally, MRv2. Mrv2 resides in the package org.apache.hadoop.mapreduce. The Oracle NoSQL Database API Hive integration classes addresses the existence of both versions of MapReduce because:
  • Hive currently employs MRv1
  • Oracle NoSQL Database API Hadoop integration classes employ MRv2
  • MRv1 and MRv2 are source incompatible

To support both MRv1 and MRv2, the Oracle NoSQL Database API Hive integration classes have subclassed the MRv1 classes to the appropriate MRv2 classes. In essence, the InputFormat from the org.apache.hadoop.mapred is a subclass of the TableHiveInputFormat from oracle.kv.hadoop.hive.table.

Note:

As the Oracle NoSQL Database Hadoop integration classes do not currently support writing data from a MapReduce job into an Oracle NoSQL Database store, the classes specified here for Hive integration do not support queries that modify the contents of a table in a store.

For more details, see oracle.kv.hadoop.hive.table, Apache Hadoop API, and Hive API.