Nonrelational NoSQL databases is the correct choice for these scenarios.

NoSQL databases scale horizontally and handle very high ingest rates while supporting flexible schemas which makes them a good fit for IoT telemetry. They also support document and key value models that work well for online retail catalogs and marketing data where the schema evolves frequently and for multiplayer gaming and web and mobile applications that need low latency and fast lookups.

Cloud Bigtable is a powerful NoSQL wide column database that excels at large scale time series and low latency workloads. It is a specific product optimized for those patterns and it is therefore a narrower choice than the general class asked for, and it may not be the best fit for transactional retail operations or workloads that require rich secondary indexes.

Relational database systems are built for strong consistency, complex joins, and structured schemas which makes them ideal for transactional workloads. They are not the best general fit for highly scalable, schema flexible, or extremely high ingest scenarios like IoT telemetry and some large scale marketing or gaming workloads.

Cloud Spanner is a globally distributed relational database that provides horizontal scale with SQL semantics and strong consistency. It is a specialized relational solution and therefore not the general NoSQL class the question asks for even though it can handle some of the scale and consistency needs of these applications.

Full AWS Practitioner Certification Question