Abstract

A node identification system is described for use in a computer system in which the various components of the system are interconnected via nodes on a communications bus. Once the topology of the nodes has been resolved into an acyclic directed graph, each node may be assigned a non-predetermined unique address. Each node having a plurality of ports has an apriori assigned priority for port selection. Each child node connected to a parent is allowed to respond in the predetermined sequence depending upon the port through which it is connected to its parent. Each node in the graph will announce its presence according to its location in the graph. Each receives an address incremented from the previous addresses assigned, thereby insuring uniqueness. The same mechanism may be implemented to allow each node in turn to broadcast information on the bus concerning the parameters of its local host. Likewise, additional information may be conveyed from each node concerning connections to other nodes thereby allowing a host system to generate a map of the resolved topology including any information about disabled links which may be used for redundancy purposes.

Other References

• Bowie, "Distributed Operating Systems," Abstract, Computer Science Conference, Feb. 20, 1975, Washington, D.C., p. 24
• Li, et al., "Grammars for Rooted Acyclic Directed Graphs and Their Parsing," Transactions of the IEICE, vol. E71, No. 4, Apr. 1988, Tokyo, Japan
• Lin & Wu, "Design of Configuration Algorithms of Commonly-Used Topologies for a Multiprocessor-Star," Proceedings of the 1985 International Conference on Parallel Processing, Aug. 23, 1985, pp. 734-741, Pennsylvania, U.S
• Michael Teener; "A Bus on a Diet--The Serial Bus alternative an introduction to the P1394 High performance Serial Bus"; IEEE 1992, pp. 316-32