Storj

Use it when

• Durability through diversity appeals: objects are erasure-coded into 80 pieces across independently operated nodes worldwide, any 29 of which rebuild the file.
• Egress-heavy sharing fits the tiers: transfer out is free up to stored volume on the live tiers.
• Parallel retrieval from many nodes suits the workload, such as large-file distribution.
• Standard S3 tooling must work; the hosted gateway is compatible, and the native uplink protocol is there for end-to-end encryption.

Think twice when

• Data residency is strict; distribution is global by default, and pinning means geofencing to the EEA or SOC 2 nodes rather than picking a region.
• Compute must sit near the data; there is no colocated analytics ecosystem, and engines read across the network.
• You planned to self-host because of the open license; the practical product is the managed network, and running a private one is possible but rare.
• Event-driven pipelines matter: notifications go only to Google Pub/Sub, and only with Storj-managed encryption.

How it runs

A managed service over a network of independent node operators, coordinated by Storj's satellites. Clients use the S3-compatible hosted gateway or the uplink protocol; the distribution is invisible in normal use beyond parallel reads. Object Lock covers compliance mode and legal holds, with governance mode incomplete.