Object Storage Can Act Like a Database Only Under These Four Strict Conditions

Object storage is not a general database replacement, but it can serve a database-like role for workloads that meet a narrow set of requirements. A real-world case involving Ampbase and Tigris illustrates how developers avoided running a traditional database by carefully rebuilding four guarantees: uniqueness, transaction-like mutation, indexing, and history. These guarantees can be approximated using two primitives that some object storage services provide — strong read-after-write consistency and conditional writes. However, this approach only holds when data is low-volume, append-only, cleanly partitioned, and accessed via predictable key lookups. If any of those assumptions break down — such as high contention or cross-object transactions — object storage forces developers to rebuild a database poorly.