CPUscheduling

CPUScheduling, commonly written as CPU scheduling, is the set of techniques used by an operating system to decide which process should run on the central processing unit at any given time. The goal is to maximize overall system performance and responsiveness while ensuring fair access to CPU time. Scheduling decisions occur when a process becomes ready, blocks on I/O, or when a running process is preempted.

A scheduler selects a process from the ready queue and transfers control to it, an operation that requires a context switch. Scheduling is influenced by workload characteristics, such as CPU-bound vs I/O-bound processes, and by the available hardware resources. Preemption is common in time-sharing systems and reduces the waiting time of interactive processes, at the cost of context-switch overhead.

Common scheduling criteria include CPU utilization, throughput, average turnaround time, average waiting time, response time, and fairness. Some criteria conflict, so systems balance them and may introduce aging to prevent starvation.

Common algorithms: First-Come, First-Served (non-preemptive, simple but can yield long waits); Shortest Job First / Shortest Remaining Time (minimizes average wait but requires future burst estimates); Priority scheduling (preemptive or non-preemptive; can cause starvation without aging); Round Robin (preemptive with fixed time quantum, good for interactivity); Multilevel queues and Multilevel feedback queues (hierarchical schemes with different policies). Real-time systems may use Rate Monotonic or Earliest Deadline First, which provide guarantees for hard deadlines.

In practice, OS implementations combine strategies and tune parameters; performance depends on workload, burst time estimates, and the cost of context switches. Starvation and aging issues require mitigation.