DataNodes

DataNodes are the storage nodes in a distributed file system such as the Hadoop Distributed File System (HDFS). Each DataNode runs on a server with local storage and stores portions of the system's data blocks. DataNodes serve data to clients and to other DataNodes during replication, while the NameNode maintains the namespace and block mapping.

DataNodes store blocks on local disks and report storage status to the NameNode via heartbeats and block reports. The NameNode uses this information to determine block locations, enforce replication, and detect under-replication. During reads, clients obtain block locations from the NameNode and fetch data from the appropriate DataNodes. Writes proceed through a pipeline of DataNodes coordinated by the NameNode; data is written to the first node and streamed to replicas to satisfy the replication policy.

DataNodes perform local integrity checks with checksums and store blocks according to the configured replication factor. If a DataNode fails, the NameNode marks blocks as under-replicated and triggers replication to other DataNodes to preserve redundancy. Operational considerations include capacity planning, network bandwidth, and hardware reliability. DataNodes typically run on commodity hardware and may be deployed across racks to improve throughput and fault tolerance. They are a core component of HDFS, distinct from the NameNode, which handles metadata and namespace management.