A systems administrator is configuring performance alerts for a critical application server with an 8-core processor. The administrator wants to be notified of a potential CPU bottleneck before it severely degrades the user experience. While overall CPU utilization is a key metric, it does not always tell the whole story. Which of the following metrics provides the MOST direct indication that the CPU cannot service all incoming requests in a timely manner?
The correct answer is Processor Queue Length. System\Processor Queue Length (or the run-queue length in Linux) shows how many runnable threads are waiting because every logical CPU core is busy. On multiprocessor systems you divide the counter by the number of cores; a sustained average higher than about two threads per core-for an 8-core server, roughly more than 16-signals that the processor is saturated and requests are backing up.
Context Switches/sec measures how often the scheduler swaps threads on or off-CPU. A high rate adds overhead but does not indicate how many threads are waiting.
Interrupts/sec counts hardware interrupts serviced by the CPU. Although frequent interrupts can raise CPU usage or reveal driver issues, it does not quantify application work queued for execution.
DPCs Queued/sec tracks deferred procedure calls that run after an interrupt. Elevated values often point to driver or I/O problems, not to general CPU scheduling backlog.