Supercomputers in The Books Store

Early high-performance computers (supercomputers) were huge, expensive, and located in computer centers, and thus limited to use by a few researchers. The increase of power and decrease in the cost of computer processors have led to a reduction in the cost of single high-performance computers, but they are still used by few organizations. On the other hand, the low cost of personal computers and workstations today enables each researcher to have his or her own computer facility, limited only by the capacity of the equipment. Such machines are often networked together, so that users can do remote log-ins and share files and public facilities in various ways.

The concept of virtual machine, a dynamic collection of (potentially heterogeneous) computational resources managed as a single parallel computer, has revolutionized distributed computers by allowing their use as a high-performance system. Nowadays portable parallel distributed software systems are available that enable a collection of heterogeneous computers to be used as a coherent and flexible concurrent computational resource. This software can integrate these machines and utilize their unused cycles to obtain a reasonably powerful computer system.

As the power of computers has grown, the complexity of the problems which can be solved by them as grown. The application of high-performance computing to numerically intensive problems in engineering raises several new issues that did not arise with standard computing. New algorithms and codes are required in order to exploit effectively the power of these new computer architectures, as programs suitable for conventional computers are likely to achieve very modest improvement of performance on high-performance computers. Some of these new techniques are addressed in the papers within this book, in particular those related to the concept of virtual machines.