Understanding Distributed Storage: File, Object, Block, and Key‑Value Systems Explained

Distributed file storage is the core form of distributed storage, targeting file read/write scenarios. Data is stored as files across nodes, providing a unified file system view. The most typical example is HDFS (Hadoop Distributed File System). Its core architecture includes a NameNode (metadata server) and DataNodes (storage nodes), suitable for massive file storage, high‑throughput sequential reads/writes, and good scalability.

Distributed object storage treats each object as the smallest management unit, storing data and metadata together. Objects are accessed via a globally unique identifier. Common implementations include Ceph. This decentralized architecture has no single metadata node; all nodes can handle requests. It is suitable for internet applications, cloud storage, backup/archive, CDN, and especially for massive unstructured data such as images, videos, and logs.

Distributed block storage provides raw storage devices in fixed‑size blocks, similar to disk partitions, which upper‑layer file systems or databases manage directly. Clients combine blocks through a virtualization layer to present a block device interface like a hard disk. It is ideal for high‑performance, low‑latency databases and virtual machine storage; typical implementations include Ceph RBD.

Distributed key‑value storage stores data as key‑value pairs or column families, emphasizing low latency and high concurrency. Typical examples are Redis (in‑memory KV), Cassandra, HBase, Dynamo, and TiKV. The data model is simple: Key and Value. Elastic scaling relies on consistent hashing and similar algorithms, allowing easy node addition or removal with automatic data migration and load balancing.