Discovery Guides Areas


Grid Computation: the Fastest Supercomputer in the World
(Released November 2006)

  by Chao-Hsu Yao  


Key Citations

Web Sites





A supercomputer is a computer designed to do large-scale and complicated computation by using many CPUs (Central Processing Units) simultaneously to perform parallel processing. Grid computation comes from the idea of super computation; it uses the internet to connect computers worldwide, to create a virtual supercomputer. Grid computation saves money and space, while a supercomputer is costly and wastes space. However, security is a concern because sharing resources always increases the risk of a computer virus or worm.


man in computer room
The ENIAC (Electronic Numerical Integrator and Computer), the first large-scale general-purpose electronic computer, in 1946

It has been almost 80 years since the first computer was invented. It was able to perform digital calculation and operated with many large sized vacuum tubes. This old computer wasted a lot of space (normally two rooms) and could only perform simple calculations. With the invention of small transistors to replace large vacuum tubes, and the progress of semiconductor manufacturing technology, the size and processing speed of the computer has greatly improved, as circuits are etched on a chip less than one micron wide. Today a personal digital assistant (aka PDA or palmtop) is the same size as a calculator with almost the same capability as a desktop personal computer, which is lighter, uses less electric power, and is less costly than a large computer.
sleek contemporary computer
Blue Gene, the fastest supercomputer in the world, can do 135.5 trillion calculations a second
♦Parallel Computation
Engineers have been seeking ways to minimize computer size and speed up processing since the invention of the first computer. Therefore, central processing unit (CPU) design has long been important in Materials Science and Engineering, Electronics Engineering, and Computer Engineering. It has been a challenge to make the CPU smaller and smaller by improving manufacturing technology. One major problem is that the heat generated from a CPU increases as its size decreases, forcing processing to slow down. To solve this size-and-speed issue, parallel processing, which uses multiple CPUs to perform parallel computations and decreases processing time, was invented. A supercomputer, normally containing thousands of CPUs, has been used to perform complicated calculations and simulation.
♦Grid Computation
Building a supercomputer facility is a huge and costly undertaking. In the early 1990s, the distributed computing system was invented at the University of California, Irvine, which connected several personal computers to perform parallel computation, instead of building an expensive supercomputer. With the invention of the internet, which connects millions of computer in the world through servers using a special protocol (aka Internet Protocol, or IP,) grid computation was invented based on the design of the distributed computing system, using the internet's powerful communication ability to connect online computers to perform parallel computation.
Kaariboga Mobile Agents
♦Comparing the Supercomputer with Grid Computation
A supercomputer has one operation system, which must be able to support parallel computation, while grid computation operates under many different operation systems, which do not need to be parallel computation supported. With grid computation, several internet protocols must be followed for communication between the computers. However, a supercomputer is just ONE computer with multi-CPU, while grid computation uses MULTI-computers connected to each other through internet protocol to simulate a supercomputer's function.

Obviously, a supercomputer with 1000 CPU's runs faster than a thousand connected computers with single CPU, because each CPU running grid computation spends extra time handling internet protocol.

Go To Theory

© Copyright 2006, All Rights Reserved, CSA

List of Visuals