I'm building a system that needs files to be accessible from multiple data centers. What storage architecture should I use?

If you need files to be accessible from multiple data centers, the usual choice is a distributed, replicated storage architecture rather than a single centralized file server.

Good options

• Distributed file system: e.g. CephFS, GlusterFS, Lustre (more HPC-oriented)
• Object storage with a shared API: e.g. S3-compatible storage across regions
• Cloud-managed multi-region storage: e.g. AWS, Azure, or GCP services with cross-region replication

What to use depends on your access pattern

• Shared POSIX-style file access: use CephFS or GlusterFS
• Large-scale app storage / blobs / backups: use object storage
• Low-latency access from each DC: use local replicas in each DC with synchronization/replication between them

Important design considerations

• Replication for redundancy and availability
• Consistency model: strong consistency vs eventual consistency
• Latency between data centers
• Conflict handling if multiple sites can write the same files
• Failover strategy if one DC goes offline

Practical recommendation

For most modern systems:

• Use object storage if you don’t need a traditional filesystem
• Use CephFS if you need a real shared filesystem across sites
• Deploy replicas in each data center and avoid a single central storage point

If you want, I can recommend a specific architecture based on:

1. read/write pattern
2. file size and volume
3. consistency requirements
4. whether you need POSIX file semantics