Hadoop

Hadoop is an open-source framework for distributed storage and processing of large data sets across clusters of commodity hardware. It is designed to scale from single servers to thousands of machines, handling petabytes of data. The core components are the Hadoop Distributed File System (HDFS) for storage, and the MapReduce processing model, together with Yet Another Resource Negotiator (YARN) for resource management. While originally built around MapReduce, Hadoop now supports a variety of data processing engines that run on the same ecosystem.

History: Hadoop originated from research at Yahoo and the Google papers on distributed storage and processing. It was created by Doug Cutting and Mike Cafarella and named after Cutting's son’s toy elephant. The project was open-sourced and joined the Apache Software Foundation in 2006. Over time, Hadoop evolved from a two-component system to a broader ecosystem, with Hadoop 2.x introducing YARN to separate resource management from job execution.

Architecture: HDFS stores data on a cluster of data nodes, managed by a centralized NameNode. Data is split into blocks and replicated to provide fault tolerance. MapReduce provides a model for parallel data processing by distributing work across nodes; YARN handles cluster resource scheduling and monitoring. The ecosystem includes tools such as Hive for SQL-like queries, Pig for dataflow scripting, and HBase for columnar storage, enabling a range of analytics tasks.

Usage and scope: Hadoop is well-suited for batch processing of very large data sets using inexpensive hardware. It emphasizes data locality, where computation is moved to where data resides. Modern deployments often integrate Hadoop with other engines such as Apache Spark, enabling faster in-memory processing, while some organizations migrate to cloud-based storage and processing services. Hadoop remains a foundational layer in many big data architectures.