On-premise object storage
Amazon S3 (2006) defined the object storage standard interface. Thousands of applications use it via SDKs. Replicating an S3-compatible service on-premise, however, requires complex distributed storage (Ceph, OpenStack Swift).
MinIO, founded in 2014 by Anand Babu “AB” Periasamy (previous Gluster co-founder), proposes a pragmatic approach: single Go binary speaking S3 API, with native erasure coding for redundancy, I/O performance optimised for NVMe.
The reference release RELEASE.2018-04-30T00-24-51Z (April 2018) consolidates production maturity. Original licence Apache 2.0, then AGPL v3.
Features
- S3-compatible API — drop-in replacement for Amazon S3 in clients
- Erasure coding (Reed-Solomon) — efficient redundancy vs. replica
- Performance multi-GB/s in published benchmarks
- Single binary — zero external dependencies, deploy in seconds
- Distributed mode — cluster up to 32 nodes per instance
- Federated — multiple clusters as single namespace
- Object locking — bucket WORM for compliance
Scenarios
- Backup target for S3-speaking products (Veeam, Restic, Duplicati)
- On-premise data lake storage for Spark, Presto, Trino
- AI/ML datasets — storage for models and training data
- Multimedia — CDN origin, file uploads
- Private cloud — part of replaceable OpenStack stack, or OpenShift, Rancher
In the Italian context
MinIO adopted in:
- Media/broadcasting — asset management
- AI/ML teams — internal dataset storage
- Hospitals/PA — clinical record backup
- Tech SaaS — private cloud S3 alternative
References: MinIO (founded 2014), Anand Babu Periasamy. RELEASE.2018-04-30. Licences: Apache 2.0 → AGPL v3 → SSPL (2021). S3-compatible API. Reed-Solomon erasure coding. Alternatives: Ceph RGW, SeaweedFS, Garage.