Operating Systems has a reputation for density — and a secret: it’s the core paper with the most guaranteed numerical questions, and numericals are the most learnable marks in any exam. Scheduling problems, page-replacement counts and banker’s-algorithm tables repeat with remarkable loyalty. Decode the families, drill the numericals, and OS becomes a scoring subject. (Map to your regulation’s units; confirm with recent papers.)
The repeating families
- Foundations & structures: OS functions/services, system calls, kernel structures — steady Part A and comfortable either-or 13-markers.
- Processes & threads: process states and the PCB (a diagram staple); threads vs processes; and the numerical king — CPU scheduling: given processes with arrival/burst times, draw the Gantt chart and compute waiting/turnaround averages for FCFS, SJF (both flavours), priority and round robin. Some scheduling numerical appears in essentially every OS paper.
- Synchronization & deadlocks — the concept heart: the critical-section problem; semaphores and the classic problems (producer-consumer, readers-writers, dining philosophers — writing the semaphore solution is a hall-of-fame 13-marker); then deadlocks: the four conditions (2-mark evergreen), resource-allocation graphs, and the Banker’s algorithm numerical — safe-sequence computation on a given matrix set, one of the most repeated questions in the subject’s history.
- Memory management: contiguous allocation and fragmentation; paging and segmentation (address-translation numericals: logical→physical with page tables); virtual memory and demand paging; and the second numerical king — page-replacement counting: given a reference string and frame count, count faults under FIFO, LRU and Optimal (plus Belady’s anomaly as the garnish 2-marker). Thrashing rounds out the unit.
- Storage & file systems: file allocation methods, directory structures, free-space management, and the third numerical family — disk-scheduling head-movement totals (FCFS, SSTF, SCAN, C-SCAN on a request queue). I/O basics and RAID levels serve Part A; per-regulation case-study tails (Linux internals where included) supply either-or options.
The examiner’s lens — the numerical discipline
OS valuers award numericals stepwise: the Gantt chart itself carries marks, the per-process table carries marks, the averages carry marks — so show every layer, and state assumptions (tie-breaking order, whether context-switch time counts) explicitly when the question is silent; valuers respect declared assumptions and punish silent ones. For theory 13-markers, the winning shape is definition → mechanism → diagram → example: a readers-writers answer with the semaphore code and two lines of explanation beats two pages of prose. Practise until each numerical type takes you under 15 minutes — exam-hall time in OS is lost almost entirely to slow scheduling arithmetic.
The industry echo
OS is a fixture of technical interviews (processes vs threads, deadlock conditions and paging are asked at every service and product company — our interview-basics post carries that angle) and the conceptual base for everything from cloud/containers to debugging production systems. The concurrency intuition you build on dining philosophers is the same one that later explains real race conditions.
The 2-week subject plan
- Days 1-2: frequency-map; collect every numerical from the last 5 papers into one problem set.
- Days 3-5: numerical bootcamp — daily: two scheduling problems, one Banker’s, one page-replacement, one disk-scheduling, timed.
- Days 6-9: concept families one per day — synchronization classics (write the solutions from memory), deadlocks, memory management theory, file systems.
- Days 10-11: 2-mark sweep (states, conditions, definitions, RAID/fragmentation one-liners).
- Days 12-14: timed full paper → autopsy → light revision of your numerical templates. Sleep.