Think BIG or Die: Envisaging the End of System Failures

摘要

Information is an essentially distributed resource in nature. It is not fully contained in central units but thoroughly split into a myriad of different parts or elements widely spread in space and time. As a result, natural systems can seldom handle all the available information. Nonetheless, such a limitation does not prevent natural organisms and ecosystems from evolving; on the contrary, it fosters competition and, ultimately, ensures life survival. Recent advances in neurosciences have shown that even one of the apparently most centralized systems, a mammal's brain, can hardly be regarded as such, but as a highly functional distributed neuronal system [2, 3]. Efficiently exploring, actively and selectively searching the surrounding environment for the most relevant information becomes a sign of intelligence and environmental fitness. Perhaps tellingly, most of the computer networks and databases are still built upon strongly centralized hierarchies. Centralized systems do work well for most of the intended purposes on a small scale. As size increases, distributed systems outperform centralized ones. But the management of a decentralized network poses new challenges that we are just beginning to address. In this paper we offer some hints and provide a description of the main characteristics that define this new paradigm of computer communities and network information systems, listing the benefits and drawbacks in computer science.