What is object storage?
Short answer: Object storage keeps each datum as an object with an identifier and metadata, accessed by API. It fits when the corpus has to grow and last; not when the workload expects a POSIX tree.
Definition
Object storage answers an operational question: how to keep a growing corpus without treating it as a directory tree that many workers have to lock. Each object is identified, versioned and read by API. It does not replace high-performance parallel storage when the workload needs POSIX and concurrent reads of the same dataset.
Why it matters
Datasets, embeddings, copies and archives grow faster than capacity plans. A file system designed for low latency is not the same problem as a repository that must last and grow.
Without a clear object layer, performance storage ends up acting as archive, or the archive tries to serve high-throughput parallel reads. Both fail in their own way.
What it is not
It is not a POSIX file system. It does not guarantee the same locks, paths or semantics that a parallel-file pipeline expects.
It is not backup. A bucket can keep versions; that is not a recovery procedure, and it is not cyber-recovery immutability merely because it is object storage.
In the Azuba model
Data & AI Infrastructure is Azuba’s solution. Object covers corpus, archive and lifecycle, distinct from high-throughput parallel file. In the published architectures, Scality covers object; WEKA covers parallel file.
Continue
Does object storage replace file storage?
Not necessarily. File access fits when the pipeline expects POSIX and parallel reads. Object access fits when the corpus grows, is archived or is accessed by API. Many production estates end up with both.
Is object storage the same as backup?
No. It can be a data destination; it does not define RPO, RTO or a restore procedure. Data Protection & Resilience is a different solution.
What does Scality cover?
The capacity object tier in Data & AI Infrastructure. WEKA covers parallel file. Fit is assessed in the specific environment.