Metacomputing. New direction in high performance computing

The primary goal of metacomputing is to integrate different computers, with different capabilities, into one unit, one large virtual computer (a metacomputer) capable of solving problems of unprecedented complexity. The whole task is controlled from microprocessor-based workstations, which also take care of all scalar (non-vectorized, non-parallelized) computations. The results of computation obtained so far are immediately displayed on several high-end graphic workstations, allowing different researchers to simultaneously inspect different aspects of the whole complex computation. A metacomputer, as a representative of the metacomputing paradigm, allows simplified access to centralized services at geographically different sites, joining them into a virtual parallel supercomputer. A personalized computing environment may be freely moved between researchers, increasing both their mobility and the efficiency of the computing resource utilization. Also, the metacomputing paradigm increases the efficient use of computers and their installed software (local mastery in some software packages is becoming available to a much larger community). Last but not least, the metacomputer enhances robustness-users at an individual site may continue in their work even in case of the failure of their local node.