Classification of Operating Systems

CSS Optional — Paper-II, Section-A · P-II.III.I (cont.)

Classification questions are graded on correct examples + correct trade-offs, not on the taxonomy alone. Never leave a category without at least one named real example.

Classification by Function/Design

Type	Defining trait	Example	Key trade-off
Batch OS	Jobs queued, no user interaction during execution	Early mainframe job schedulers	High throughput, poor turnaround for any one job
Time-Sharing OS	Fast preemptive CPU switching gives multiple interactive users the illusion of a dedicated machine	UNIX, multi-user Linux	Good responsiveness, more scheduling/context-switch overhead
Real-Time OS (RTOS)	Guarantees a response within a strict deadline	Pacemakers, flight-control systems (hard); video streaming (soft)	Predictability prioritized over average throughput
Distributed OS	Manages a group of physically separate, networked computers and presents them as a single coherent system	Google's internal cluster OS, Amoeba	Scalability and fault tolerance vs. complexity of coordination
Network OS	Each machine keeps its own local OS/autonomy but the NOS provides services (file sharing, printing) across the network	Novell NetWare, Windows Server	Simpler than a true distributed OS, but less transparent — users must know which node hosts a resource
Mobile OS	Optimized for touch input, battery/power constraints, sensor integration	Android, iOS	Aggressive power management can throttle background work
Embedded OS	Runs a single dedicated application on resource-constrained hardware	Firmware in a washing machine, router OS	Minimal footprint, but little to no multitasking flexibility

Hard vs Soft Real-Time (a favourite sub-distinction)

Hard real-time: missing a deadline is a total system failure — e.g., an airbag controller or a pacemaker. No missed deadline is ever tolerable.
Soft real-time: an occasional missed deadline degrades quality but the system keeps working — e.g., a video call frame arriving late just causes a visible stutter.

Distributed OS vs Network OS (the pair examiners most often confuse candidates on)

A Distributed OS gives transparency: a user runs a program without knowing or caring which physical machine executes it. A Network OS gives no such illusion: each machine is independently managed, and the user must explicitly connect to (or mount/map) a remote resource to use it (e.g., mapping a network drive).

Clustered Systems (a classification type often left out — worth adding to any "classify OS" list)

Clustered systems gather multiple complete computers (each with its own OS) tightly linked together, sharing storage and connected via a fast local network, to work as a team providing high performance and/or high availability.

Asymmetric clustering: one machine is in hot-standby mode, monitoring the active server, and takes over only if the active server fails.
Symmetric clustering: two or more machines all run applications and monitor each other simultaneously — more efficient since no machine sits idle as a pure standby.

Clustering differs from a true Distributed OS in that each node still runs its own independent OS instance rather than the cluster appearing as one seamless system.

Multiprocessor (Parallel) Systems — a design axis orthogonal to the table above

Not strictly a separate "type" of OS in the classification table, but frequently confused with it: a multiprocessor system has two or more CPUs in close communication, sharing the same physical memory and I/O, managed by one OS.

Symmetric Multiprocessing (SMP): every CPU runs an identical copy of the OS and any CPU can run any process — the common modern model.
Asymmetric Multiprocessing (AMP): one "master" CPU controls the system and assigns work to other "slave" CPUs — simpler, but the master can become a bottleneck.

🎯 CSS Exam Angle

Classification alone is rarely worth full marks. Pair it with a scenario: "which type of OS would you recommend for X, and why?" — e.g. a hospital ICU monitor (hard RTOS, deadline correctness over throughput) vs. a university's shared computing lab (time-sharing, fairness across many interactive users). If the scenario clearly involves several tightly-linked machines acting as a team, don't forget clustering as a valid answer alongside distributed/network OS.