A SysOps admin maintains a legacy application that is CPU-bound and can only scale by increasing instance size. It currently runs on a single t3.large EC2 instance and CPU utilization spikes to 90% with high latency after a few minutes. Which modification will most directly reduce the CPU-related performance bottleneck?

Switching to a compute-optimized instance type (e.g., C-series) will directly address the CPU bottleneck. These instance types are designed for CPU-intensive workloads and offer significantly higher CPU performance compared to general-purpose instances like t3.large, which are burstable and can struggle under sustained high CPU load. Converting the EBS volume to Provisioned IOPS would improve disk I/O, but the problem states CPU utilization is the bottleneck, not disk I/O. Launching additional t2.large instances won't help because the application cannot scale horizontally; it requires a larger single instance. Buying Reserved Instances is a cost-saving measure, not a performance improvement.