Binary compatibility here means that the compiled binaries should be able to run without any modification on the new framework.
For those Hadoop 1.x users who use the org.apache.hadoop.mapred APIs, they can simply run their MapReduce jobs on YARN just by pointing them to their Apache Hadoop 2.x cluster via the configuration settings.
They will not need any recompilation. All they will need to do is point their application to the YARN installation and point HADOOP_CONF_DIR to the corresponding configuration directory. The yarn-site.xml (configuration for YARN) and mapred-site.xml files (configuration for MapReduce apps) are present in the conf directory.
Also, mapred.job.tracker in mapred-site.xml is no longer necessary in Apache Hadoop 2.x. Instead, the following property needs to be added in the mapred-site.xml file to make MRv1 applications run on top of YARN:
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
For those Hadoop 1.x users who use the org.apache.hadoop.mapred APIs, they can simply run their MapReduce jobs on YARN just by pointing them to their Apache Hadoop 2.x cluster via the configuration settings.
They will not need any recompilation. All they will need to do is point their application to the YARN installation and point HADOOP_CONF_DIR to the corresponding configuration directory. The yarn-site.xml (configuration for YARN) and mapred-site.xml files (configuration for MapReduce apps) are present in the conf directory.
Also, mapred.job.tracker in mapred-site.xml is no longer necessary in Apache Hadoop 2.x. Instead, the following property needs to be added in the mapred-site.xml file to make MRv1 applications run on top of YARN:
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
