IOScheduling

IO scheduling, or disk scheduling, is the part of the operating system that determines the order in which storage I/O requests are serviced. The objective is to improve performance, reduce latency, and ensure fairness among processes while avoiding starvation. Each storage device has a request queue managed by the IO scheduler; the chosen algorithm decides which pending request to execute next based on factors such as head position, device geometry, and service quality requirements.

Common scheduling strategies include first-come, first-served (FCFS), shortest seek time first (SSTF), and the elevator family (SCAN) which moves the drive head in one direction, reversing at the end; LOOK and CLOOK are optimized variants that do not travel to the physical end unless required. N-step and FSCAN divide requests into batches to reduce long-term unfairness. Modern systems also implement deadlines, fairness controls, or quality-of-service targets to meet varying workload requirements.

In hard disk environments, scheduling aims to minimize seek distance and rotational latency, whereas solid-state drives reduce or eliminate seek-based differences, sometimes leading to simplified or hardware-driven scheduling. Operating systems expose multiple I/O schedulers or allow tunable policies; examples include CFQ, Deadline, NOOP, BFQ, and others, with behavior varying by workload and hardware.

The I/O scheduler operates within the kernel’s block layer and interacts with caching, prefetching, and RAID or other adapters. It is a core component for performance in multi-user and I/O-intensive systems, and its design reflects trade-offs among throughput, latency, predictability, and fairness.