Types of Computers - the Big Ones (Page 2 of 2)

Grid computing is an application of cluster computing. David Bader, in his paper, available as a pdf download, Cluster Computing: Applications, in The International Journal of High Performance Computing, 15(2):181-185, May 2001 , says

“The performance of an individual processor used in a high-end personal workstation rivals that of a processor in a high-end supercomputer, such as an SGI Origin, and the performance of the commodity processors is improving rapidly.”

Written in 2001, Bader is talking about state of the art computers of the time, used in clusters. The strides made in personal computers since mean the clustering he discusses in the paper used with the computers of today make possible a degree of computing power he does not imagine.

Grid computing is the use of many personal computers/workstations to solve distributed problems. Each individual computer, or node, on the grid receives a packet of information to process when the computer is not otherwise active. It actually does not matter how advanced or high-end the particular computer is, as long as it can process the project it is given. No other node is directly connected to that PC, so it does not need to have any synchronization with any other node. All information is sent back to a central computer, and it sends out a new task when a node completes a task.

Several grid computing projects have the accumulated processing power of computers in the top500 list of supercomputers. It is estimated that nearly three million home computers around the world are working on the SETI@home project. While this does not come near the speed of a top level supercomputer, or mainframe, it is the astonishing number of tasks that these networks can handle which makes them remarkable. It also is a way for personal computer users to participate in some of the most advanced scientific work on the planet, simply by allowing access to their computer when they are not using it. Other large grid computing projects are working on Alzheimer's disease and on information from the Large Hadron Collider.

A mainframe is what people usually think of when they think of the computers that were around before the era of personal computing. Mainframes share certain characteristics with servers, such as their ability to store data redundantly and their ability to keep running even though part swapping is happening. There are mainframes that have been running without failure for more than ten years. Mainframes have lots and lots of power. They can run many many operations at once, such as all the financial transactions happening at all Bank of America locations. In fact, it is primarily governments and corporations needing to answer high volumes of requests that use mainframes now.

Mainframes started out as room sized machines, and now are as small or smaller than refrigerators, but they have some specific requirements to keep them in good working order. They need a dust free temperature controlled environment. They do not, however, have the high electrical consumption that characterized old mainframes, and may use much less energy than the "server farms." A few new mainframes are still being produced. Many mainframes still use magnetic tape to back up on.

The last big computer we will look at is the supercomputer. Supercomputers are usually built to do one thing, and do it really really fast. Most of them are purpose built, as needed, and there may often only be one in the world. Supercomputers work on such things as DNA sequencing, or cryptography. They will also be involved in the new square kilometer radio telescope, once it is built.

Their speed is measured in FLOPS, or floating point operations per second, and the newest ones are mind bogglingly fast, and measured in petaflops. This gives them a theoretical speed of 1. with 15 0s after operations per second, or a thousand trillion operations per second. A very few supercomputers are available for use for general science, but most are purpose dedicated.

In the next article, we will look at all the different types of computers lumped into the category of personal computer.

Source material and corrections for this article came from Eofn Williams, Henry Scudder, Michael Scudder, Lamar Stonecypher, Michele McDonough, James Allen Johnson, Wikipedia, Seti@home, David Bader's paper archived by Georgia Tech, and Top500 site, which lists the top 10 supercomputers with links to additional material on all of them. A wonderful pictorial history of computers from the early 20th century through the 80s can be found at the Computer History Museum.