An architect asks why watsonx.data uses object storage as its primary storage layer rather than a high-performance shared filesystem. What is the strongest justification?

The strongest justification is that object storage scales capacity and durability independently of compute at low cost, which is what lets many engines share one copy of the data. Object storage offers effectively unlimited capacity, high durability and low cost, and it decouples storage from compute so engines can be added, resized or removed without moving data. That decoupling is the foundation of the multi-engine shared-data model. The claim that it provides the lowest possible per-query latency for analytical scans is incorrect because object storage is not chosen for latency; table formats, file layout, partitioning and engine caching are what make analytical performance acceptable on top of it. The claim that it enforces access control at the row level, removing the need for a governance layer, is incorrect because object stores control access at the object and bucket level. Row and column entitlement comes from the governance layer, which is a separate component precisely because storage cannot express it. The claim that it stores table metadata natively, removing the need for a catalog, is incorrect because a bucket holds objects, not table definitions. The catalog exists to supply schemas and the file sets that constitute tables. For the exam, phrase the benefit as scalable, durable, cost-effective, and independent of compute.