Virtual computer architecture for network PCs


Juhász Sándor <sanyo@sch.bme.hu>

BME, Automatizálási és Alk. Informatikai Tanszék

Charaf Hassan Phd. <hassan@aut.bme.hu>

BME, Automatizálási és Alk. Informatikai Tanszék


As we are approaching more and more to the limit, where the present integrated circuit technology it is not capable to improve the speed of computation any more, the other way of increasing performance, the application of parallelism is becoming more and more dominant. It is present on multiple level, reaching from parts of circuits to even multiplying whole computers. Its most impressive form, the Internet physically connects the computers of the world, allowing the distribution of data storing and data processing on a scale never seen before. The goal of the researches in this field is to create the computational GRID, a demand driven processing architecture extending to the whole planet.

The metacomputing systems based on various programming solutions aim to unite the computational forces of their heterogen components. Such systems also share a number of base requirements like performance, single system image, fault tolerance etc., so solutions addressing these issues also have some common characteristics. Some base components (identification, communication, synchronisation, security system, information service) are common parts of every system, only the implementations differ. Our department also pursuits the development of a parallel, distributed framework, which we would like to present in this paper. Our system makes it possible to use and to program PCs connected by the Internet as a single multiprocessor system.