Research problems of decentralized systems with largely autonomous nodes

Abstract

A currently popular systems research project is to explore the possibilities and problems for computer system organization that arise from the rapidly falling cost of computing hardware. Interconnecting fleets of mini- or micro-computers and putting intelligence in terminals and concentrators to produce so-called "distributed systems " has recently been a booming development activity. While these efforts range from ingenious to misguided, many seem to miss a most important aspect of the revolution in hardware costs: that more than any other factor, the en_ ~ cost of acquiring and operating a free-standing, complete computer system has dropped and continues to drop rapidly. Where a decade ago the capital outlay required to install a computer system ranged from $150,000 up into the millions, today the low end of that range is below $15,000 and dropping. The consequence of this particular observation for system structure comes from the next level of analysis. In most organizations, decisions to make capital acquisitions tend to be more centralized for larger capita] amounts,