Pick a scheduling algorithm and a workload, then watch the CPU timeline and the ready queue.
What the scheduler does. Several processes want one CPU. At every tick the OS picks one ready process to run. FCFS runs the earliest arrival to completion; SJF picks the shortest burst; SRTF is the preemptive version (always the least remaining time); Round Robin gives each process a fixed quantum and then moves it to the back of the queue; Priority picks the lowest priority number and may preempt.
Metrics. Turnaround = finish - arrival. Waiting = turnaround - burst. Response = first run - arrival. Utilisation = busy ticks / total ticks. A context switch is counted when the CPU changes from one process directly to another.
Known results. For bursts 24, 3, 3 arriving together FCFS gives average wait 17 (the convoy effect) while SJF gives 3; SJF/SRTF are optimal for average waiting time but can starve long jobs. A short Round Robin quantum improves response but adds context switches.
Honest limits: one CPU, discrete time, bursts known in advance (real OSes must estimate them), context switches cost zero time, no I/O waits, memory or file systems; the description's memory and file-system topics are not modelled here.