The correct answer is It retains frequently used data entirely in memory for immediate retrieval.

This option is correct because an in memory cache keeps data in RAM so reads are served without disk access and that removes the latency associated with disk operations. Memory access is orders of magnitude faster than disk access and that is the primary reason a cache can return data much faster than a disk based relational database.

In addition caches perform simple key value lookups and avoid the overhead of complex query planning, transactional durability, and frequent writes to persistent storage. Those design differences reduce CPU and I/O work which further improves latency compared with a relational database that is optimized for consistency and persistence.

It stores data on high performance NVMe solid state drives is incorrect because typical caching systems like Memcached store data in RAM rather than on SSDs. Even NVMe SSDs have higher latency than RAM and the cache advantage comes from memory resident data.

Cloud Memorystore is incorrect because it is a product name for a managed caching service and not an explanation of why an in memory cache is faster. While Cloud Memorystore can provide in memory caching, the performance reason is the data being kept in memory rather than the service name.

The cache runs on the same physical host as the application is incorrect because caches do not need to be co located to be fast and the defining factor for speed is avoiding disk I/O by serving from RAM. Co locating a cache can reduce network latency but it is not the fundamental reason an in memory cache is faster than a disk based database.

Full AWS Practitioner Certification Question