Technical overview of the intelligent file store

摘要

Large knowledge-based systems tend to perform poorly on conventional computer architectures. In order to improve performance, a class of machines called ‘knowledge manipulation engines’ has been proposed. This paper analyses the functional requirements of such machines, in terms of both memory- and processing-dominated activity. The Intelligent File Store (IFS) is introduced as a novel knowledge manipulation engine which addresses these functional requirements. The IFS architecture is described, together with examples of knowledge represenetation formalisms which it supports. The performance of the IFS hardware is compared with conventional software solutions.