|
In this article:
- What is a Supercomputer
- Applications of Supercomputers
- Supercomputer Design
What is a Supercomputer?
Supercomputing as a term was originally used to refer to large IBM tabulators by a newspaper in 1929. Today, a supercomputer denotes a computer system that has very high processing capacity and speed and is very, very superior to the other computers of its time.
Supercomputers were first introduced in the 1960s most of which were made by Seymour Cray at the Control Data Corporation (CDC). Later, Cray started his own company under the name of Cray market and swept the supercomputing market off its feet with his innovative designs. His computers (since he never used the word ‘supercomputer’ to refer to them) led the world in terms of speed, processing power and other features from 1985 to 1990. Supercomputers, as of now, are custom-built, powerful computing ‘powerhouses’. IBM and HP are important companies producing supercomputers. However, Cray Inc. exists as a separate company today as well and specializes in what it pioneered, that is, supercomputers.
A typical supercomputer is collection of many processors working in a parallel arrangement (that is, parallel computing). A supercomputer usually has 4 to 16 processors. However, supercomputers may also make use of many simple CPUs at a time. These CPUs are usually RISC microprocessors such as PowerPC from IBM. Thus, essentially, supercomputers are nothing but a computer cluster.
Applications of Supercomputers
Supercomputers are apt for providing solutions to problems that require a high amount of calculations. This includes weather forecasting, molecular modeling, simulations systems and cryptanalysis etc. Supercomputers usually find use in universities, research labs and military agencies.
Supercomputer Design
Supercomputers haven an edge over usual computers in that they are very fast and highly efficient. They are a result of complex engineering and design. Supercomputers today are usually made for a specific task and hence are custom-made for that particular purpose only. A typical characteristic of a supercomputer that differentiates it from a usual computer is its memory hierarchy which ensures that the processor always has some or the other data to process. This is different from a usual computer where there are times where the processor is ‘free’. The Input/Output systems of a supercomputer are built to work with very high bandwidth.
However, supercomputers tend to have certain issues owing to their own qualities:
- First and foremost is heat generation. Since supercomputers do a lot of processing at a time, they generate a huge amount of heat. Thus, keeping them cool is a major problem.
- Since there exists an upper limit for the speed at which data can travel within a supercomputer (that is the speed of light), there needs to be the least possible ‘cable travel’ for the data which would greatly hinder the speed of the supercomputer. It is for this reason that an ideal shape for a supercomputer would thus be cylindrical (as in the Cray supercomputers).
- Since computers process data at blazing speeds, they are able to produce a huge amount of data at a time. This becomes an issue as measured need to be taken to ensure that the data is transferred fast and stored (or retrieved) as soon as possible.
When it comes to operating systems, supercomputers originally used Unix. However, the trend today has moved towards using its variants, most prominently Linux. It is notable that supercomputer OSes do not generally have a ‘friendly’ user interface as they need to concentrate on processing. Supercomputers are usually programmed using Fortran compilers as it has proved to generate code faster than C or C++. Also, it is very suited for scientific programming. Thus, most programs that supercomputers ‘execute’ are written in Fortran. Various custom applications have been built to make use of the parallel nature of supercomputers (eg., PVM, MPI etc.)
|
