A company has an SLA of 15ms for storage latency. Given the above metrics, which of the following is the MOST appropriate change to the environment?

Adding computing and data nodes scales capacity and aggregate throughput but does not reduce per-operation latency on the existing storage array where the SLA is being missed.

Adding storage shelves increases raw capacity but does not reduce the latency of individual I/O operations already being processed by the array.

Compression reduces the physical size of data written to and read from storage media, decreasing the volume of I/O operations the array must process per request, which directly lowers effective latency. Unlike deduplication, modern inline compression adds minimal processing overhead and avoids the hash-lookup delays that deduplication introduces. This makes it the most targeted and cost-effective adjustment for an environment failing a latency SLA.

Enabling deduplication introduces additional processing for hash computation and lookup operations, which can increase storage latency rather than reduce it.