An administrator is configuring Storage I/O Control. After enabling, the administrator notices high device latency and poor storage performance. Which two actions would likely reduce latency and impro

Shared spindles across multiple datastores cause SIOC's per-datastore latency measurements to reflect combined I/O competition from all datastores, making it impossible for SIOC to correctly identify which datastore is congested and rendering its throttling decisions inaccurate.

SIOC measures device latency at the datastore level to identify congestion and arbitrate I/O shares among competing VMs. When multiple datastores share the same physical spindles, I/O activity on one datastore inflates latency readings for others, causing SIOC to throttle I/O on datastores that are not actually congested. Independent spindles per datastore ensure that SIOC's latency measurements are isolated and accurate, enabling correct congestion detection.

SIOC activates its I/O throttling when measured datastore latency exceeds the configured congestion threshold. A threshold of 15ms is close enough to realistic workload latency that SIOC engages early to redistribute I/O shares before latency becomes severe, without being so low that it throttles normal healthy traffic. This more responsive value directly reduces the high device latency the administrator observed.

A 5ms congestion threshold is far below normal storage latency for most workloads, causing SIOC to throttle I/O almost continuously and severely degrading overall storage performance rather than improving it.