Chapter 12. Real-time Data Analytics Using YARN

Hadoop is known for batch processing of data available in HDFS through MapReduce programming. The data is placed in HDFS before it can be queried for analysis. The Hadoop services execute only MapReduce jobs. The cluster resources are not fully utilized for other operations when the resources are ideal.

This is considered as a limitation for use cases that required processing of data in real time. Apache Storm and Spark are the frameworks developed for processing data in real time. These frameworks need an efficient cluster's ResourceManager. Focusing on a common solution for the preceding limitations in Hadoop, YARN evolved as a generic framework to provide resource management and application execution over a cluster. It not only allows different frameworks other than MapReduce to use the same cluster but also provides efficient scheduling algorithms to the applications running on the cluster. Frameworks such as Storm, Spark...