❓ Operating Systems - Q&A¶

Pairs with Operating Systems. Format and tag legend: Overview.

Q1. Process vs thread? A process is an isolated program with its own address space; a thread runs inside a process and shares its memory. Threads are cheaper to create and switch and communicate via shared variables; processes are isolated, so a crash in one doesn't corrupt others but they need IPC to talk. Follow-ups: When would you pick multiprocessing over threads? (CPU-bound work, fault isolation.) Easy · MAANG

Q2. What happens during a context switch, and why does it cost? The OS saves the current thread's CPU state (registers, PC, stack pointer) and loads another's so the CPU can switch execution. It's pure overhead - no useful work happens - and it also evicts cache/TLB entries, so excessive switching hurts throughput. Follow-ups: What triggers one? (time-slice expiry, blocking I/O, higher-priority thread.) Med · MAANG

Q3. What is virtual memory and why do we have it? Each process gets its own virtual address space that the OS/MMU maps to physical RAM in pages. It gives isolation (processes can't touch each other's memory) and overcommit (total virtual memory can exceed RAM, with inactive pages paged to disk). Follow-ups: What's a page fault? What's thrashing? Med · MAANG

Q4. Explain paging and the TLB. Memory is split into fixed-size pages; page tables map virtual pages to physical frames. The TLB caches recent translations so the CPU doesn't walk the page table every access - a TLB miss makes that access much slower, which is why memory locality matters. Follow-ups: What is demand paging? Med · Google

Q5. What is thrashing? When the combined working set exceeds physical RAM, the system spends most of its time paging pages in and out rather than doing work, and throughput collapses. The fix is more RAM or a smaller working set - not more swap. Follow-ups: How would you detect it in production? (high page-fault rate, low CPU utilization.) Med · MAANG

Q6. Stack vs heap? The stack holds call frames and locals, is per-thread, and is auto-freed on return (fast, limited size). The heap holds dynamically allocated objects, is shared within a process, and lives until freed/GC'd (flexible, slower, can fragment or leak). Follow-ups: What causes a stack overflow? (deep/infinite recursion.) Easy · MAANG

Q7. What are the four Coffman conditions for deadlock? Mutual exclusion, hold-and-wait, no preemption, and circular wait - all four must hold simultaneously. Breaking any one prevents deadlock. Follow-ups: Which is easiest to break in practice? (circular wait, via a global lock ordering.) Med · MAANG

Q8. How do you prevent deadlock? The most practical approach is to impose a global resource-ordering so a circular wait can't form. Others: avoidance (Banker's algorithm - rarely used), or detection + recovery (detect cycles, kill a victim). Follow-ups: Difference between deadlock prevention and avoidance? Med · Amazon

Q9. Preemptive vs cooperative scheduling? Preemptive: the OS can interrupt a running thread at any time (fair, responsive, what modern OSes use). Cooperative: a task runs until it voluntarily yields (simpler, but one misbehaving task can starve everyone) - the model behind async event loops. Follow-ups: Where do you see cooperative scheduling today? (asyncio, older systems.) Med · MAANG

Q10. Name a few CPU scheduling policies and a trade-off of each. FCFS (simple, but long jobs block short ones), Round Robin (fair/responsive, switch overhead), Shortest Job First (optimal wait, can starve long jobs), Priority (can starve low priority - fix with aging). Real OSes approximate a multilevel feedback queue. Follow-ups: What is priority inversion? Med · Google

Q11. What is a system call? The controlled entry point from user mode into kernel mode - how a program requests a privileged operation (open a file, send on a socket). It involves a mode switch, which is costlier than a normal function call, so hot paths batch syscalls. Follow-ups: Why is user/kernel mode separation important? (isolation, stability, security.) Med · MAANG

Q12. User mode vs kernel mode? User mode runs application code sandboxed with limited privileges; kernel mode has full hardware access. The boundary keeps a buggy or malicious program from crashing the machine or reading other processes' memory. Follow-ups: What causes a transition into kernel mode? (syscalls, interrupts, faults.) Easy · MAANG

Q13. How do processes communicate (IPC)? Via pipes, message queues, shared memory, sockets, or signals. Shared memory is fastest (no copy) but needs synchronization; message passing is safer but copies data. Follow-ups: Trade-off between shared memory and message passing? Med · Amazon

Q14. What is a zombie process? An orphan? A zombie has finished but its exit status hasn't been reaped by the parent, so its entry lingers in the process table. An orphan's parent died first; it gets re-parented (to init/PID 1), which reaps it. Follow-ups: How do you avoid accumulating zombies? (parent calls wait().) Hard · Google

Q15. Concurrency vs parallelism (at the OS level)? Concurrency is the OS interleaving multiple threads so they make progress (possible on one core); parallelism is threads literally running at once on multiple cores. Scheduling gives concurrency; multiple cores give parallelism. Follow-ups: Can you have concurrency without parallelism? (yes - single core, time-sliced.) Easy · MAANG

Q16. What is a mutex vs a semaphore (OS view)? A mutex allows one holder at a time (ownership-based mutual exclusion). A semaphore is a counter allowing up to N holders - useful to cap concurrent access to a pool of resources. A binary semaphore resembles a mutex but has no ownership concept. Follow-ups: Why does ownership matter? (correct release, priority inheritance.) Med · MAANG

Q17. What is memory fragmentation? External fragmentation: free memory is split into scattered gaps too small to satisfy a request even though total free space is enough. Internal: an allocated block is bigger than requested, so the slack is wasted. Compaction or better allocators mitigate it. Follow-ups: How does paging help with external fragmentation? (fixed-size frames avoid it.) Hard · Google

Q18. Why is a page size a trade-off? Larger pages mean smaller page tables and better TLB coverage but more internal fragmentation and wasted I/O when paging. Smaller pages waste less memory but need bigger page tables and cause more TLB misses. Follow-ups: What are huge pages used for? (databases, JVMs - reduce TLB pressure.) Hard · Amazon