Generations and classification of computers

The examinable point behind the table: each step packs more switches into less space, and every other improvement — speed, size, cost, reliability, heat — follows from that.

A hybrid computer contains both: an ICU patient monitor measures pulse and temperature as analog signals and then processes them digitally to display readings and raise alarms.

1. Microcomputer. Built around a single microprocessor and intended for one user at a time. Smallest, cheapest and most numerous. Includes desktops, laptops, notebooks, tablets, palmtops, smartphones and workstations. A workstation is a high-end microcomputer for engineering, design or graphics work.

2. Minicomputer. Medium-sized and multi-user, serving a department or small organisation with anything from a few to a few hundred terminals. More powerful and more expensive than a microcomputer, less so than a mainframe. Historic examples: PDP-11, VAX-8000. The category has largely been replaced by powerful servers.

3. Mainframe computer. Very large, supporting thousands of simultaneous users. Designed for extremely high transaction throughput and for continuous, reliable operation over years rather than for raw arithmetic speed. Requires a controlled environment and specialist staff. Used by banks, railways, insurance companies and tax departments. Example: IBM Z series.

4. Supercomputer. The fastest and most expensive class, built to run one enormous calculation as quickly as possible rather than to serve many users. Achieves its speed by massive parallel processing across thousands of processors. Performance is measured in FLOPS. Used for weather forecasting, climate modelling, molecular and nuclear simulation and aerodynamic design. Indian examples: PARAM series, AIRAWAT.

The key contrast to state explicitly: a mainframe is optimised for many transactions, a supercomputer for one calculation. Both are large, but for opposite reasons.

Every difference in the table traces to a single cause: the number of switches that can be placed in a given volume rose by many orders of magnitude, so the machine became smaller, cooler, cheaper, faster and more reliable all at once.

A hybrid computer contains both an analog and a digital section, using each for the part of the job it does best. The analog section measures a continuously varying physical quantity; the digital section then processes, stores and displays that measurement.

Standard example — an ICU patient monitor.

1. Analog section. The patient's temperature, pulse rate, blood pressure and blood-oxygen level are continuous physical quantities. Sensors convert each into a continuously varying electrical voltage. Only an analog device can pick up a continuous signal in the first place.

2. Analog-to-digital conversion. The voltage is sampled and converted into a number.

3. Digital section. The numbers are displayed as readings, plotted as a trace, stored as a record of the last several hours, compared against safe limits, and used to trigger an alarm when a limit is crossed.

Why both are needed. The quantities being observed exist in the physical world as continuous signals, so measurement must be analog. But storing a history, comparing against thresholds, raising an alarm and printing a report are all discrete decisions on stored data — work only a digital computer can do accurately and repeatably.

Other examples: a petrol pump (analog flow measurement, digital price calculation), and industrial process controllers.

They are large for opposite reasons, and each is poor at the other's job.

What a mainframe optimises. Throughput of many small independent transactions, and uninterrupted availability. A booking system might handle tens of thousands of requests per second, each of which does almost no arithmetic — read a record, check a seat, write a record. The design effort goes into:

• very high input/output bandwidth, because the work is dominated by moving records, not computing
• transaction integrity, so a booking is never half-completed
• redundancy and hot-swappable parts, so the machine runs for years without stopping
• serving thousands of concurrent users fairly

What a supercomputer optimises. The wall-clock time of one enormous calculation. A weather model divides the atmosphere into millions of cells and solves equations for each, repeatedly. The design effort goes into:

• the largest possible number of floating-point operations per second, across thousands of processors working on one problem
• extremely fast interconnect between processors, since they must exchange partial results constantly
• vector and parallel arithmetic units

Why they are not interchangeable.

Put the reservation system on a supercomputer and most of its arithmetic units sit idle while it waits on disk and network, and it has neither the transaction machinery nor the year-round reliability the job requires — an expensive machine doing an I/O-bound job badly.

Put the weather model on a mainframe and the single calculation runs far too slowly to be a forecast; a prediction of tomorrow's rain that takes three days to compute has no value at all.

The general principle. "Powerful" is not one quantity. A machine is fast at the thing it was designed to be fast at — here, transactions per second versus floating-point operations per second — and choosing hardware means matching the design to which of those the workload actually needs.