Design Philosophy and Architecture of JuiceFS: A Cloud‑Native Distributed File System

JuiceFS is a cloud‑native distributed file system that emerged after its open‑source release in early 2021, gaining rapid community adoption. The presentation introduces its background, design goals, and the value it brings to developers.

Development of file storage – The article traces storage from the LAN era, through the Internet boom, the mobile Internet era, and finally the cloud‑native era, highlighting how each generation introduced new requirements and limitations.

Cloud‑era pain points and challenges – It compares the characteristics of traditional NAS, early software‑defined storage (HDFS, CephFS), and object storage (S3), emphasizing issues such as limited APIs, metadata performance, lack of atomic rename, and eventual consistency.

JuiceFS design philosophy – The system aims to be a cloud‑optimized file system that supports multiple scenarios, separates metadata from data, adopts a plugin‑based engine, and follows the “Simple is better” principle.

Key capabilities – JuiceFS provides service‑oriented operation, multi‑cloud support, elastic scaling, high availability, high throughput, low latency, strong consistency, and efficient handling of massive small files.

Architecture – The architecture consists of three pluggable components: a metadata engine (supporting Redis, MySQL, PostgreSQL, TiKV, SQLite, BadgerDB, ETCD, etc.), a data engine that leverages any object storage service, and a client exposing four access methods (POSIX‑compatible FUSE, HDFS‑compatible Java SDK, CSI driver for Kubernetes, and an S3 gateway).

Observability – JuiceFS includes detailed access logs and CLI/GUI tools to help developers and SREs diagnose performance issues.

Use cases – Common scenarios include Kubernetes persistent volumes, AI pipelines, big‑data workloads (e.g., Hadoop, ClickHouse, Elasticsearch), and migration of traditional NAS to the cloud for industries such as genomics, drug research, and remote sensing.