Topic 5 · Resource Management (HL)

Manages hardware/software resources. The referee between apps and hardware.

Program = file on disk. Process = running instance with its own memory, state, and PID.

NEW → READY → RUNNING → (BLOCKED ↔ READY) → TERMINATED.

Process waiting for I/O. OS runs something else meanwhile — that's why I/O doesn't waste CPU.

Save current state to PCB + load next process's state. Costs microseconds + pollutes caches.

First-Come-First-Served. Simple. Convoy effect: one long job blocks shorter ones behind it.

Shortest Job First. Optimal AVERAGE wait. Risk: long jobs STARVE; OS may not know job lengths.

Fixed time slice (e.g. 10 ms) per process; rotate. Best for INTERACTIVE systems (laptops, phones).

Highest priority runs. Best for REAL-TIME deadlines. Risk: low-priority STARVATION (fix: aging).

Real OSes combine RR + priority + aging. Linux, Windows, macOS all do this.

Shorter slice = more responsive but more context-switch overhead. Bigger = vice versa.

Each process gets its OWN linear address space; OS+MMU translate to physical RAM.

Isolation (security + stability) + Abstraction (programmer simplicity) + Capacity (run more than fits in RAM).

Fixed-size chunk, typically 4 KB. Unit of virtual→physical mapping.

A page-sized slot in physical RAM. Page tables map pages → frames.

Per-process table mapping each virtual page number to a physical frame (or 'on disk').

Memory Management Unit — hardware on the CPU that performs page-table lookups on every memory access.

Translation Lookaside Buffer — small cache of recent page-table lookups for speed.

Page not in RAM → interrupt → OS fetches from disk (~10 ms) → retry. 100,000× slower than RAM (~100 ns).

Working set > RAM → constant page faults → CPU idle + disk 100% → system unusable. Fix: free RAM.

Hardware-enforced via MMU + page tables. Process A simply CAN'T address Process B's pages. Segfault on violation.

CPU repeatedly asks 'are you ready?'. Simple. Wastes cycles for sparse events.

Device signals CPU; CPU pauses, runs ISR, resumes. Efficient for sparse / unpredictable events.

Interrupt Service Routine — SHORT OS handler that acknowledges + moves data + wakes blocked process + returns.

Interrupt Vector Table — maps interrupt number to ISR address.

Very high-rate predictable streams (10+ Gbps networking via NAPI). OR hardware with no interrupt support.

Sparse + unpredictable (keyboard, mouse, disk-read complete, network packet arrival).

Linux convention: ISR does minimum (top); awakened kernel thread does heavy work (bottom).

Keep SHORT. Long ISRs delay other interrupts, can cause drops + unresponsive system.