These workloads combine the storage decisions you have already used. A correct component list is only the beginning. You need to tell the board how reads are routed, how a busy key is split, and how a cache learns about changes. Each task gives a new business constraint.
Work from the operation that must remain correct during a failure. A profile editor needs a session freshness rule; a busy event tenant needs distribution inside its own key; a disappearing cache needs a protected storage fallback. You'll also calculate capacity with the units supplied in the prompt. Explain the trade-off through your selected reason. A design that shifts cost to reads can be valid, provided the read workload can afford that cost.