Fail-Safe Local Area Networking Using Channel Redundancy.

Abstract

Local area networks represent a major development in data communications. The ring topology is the most popular of the local area networks. Its inherent problem, however, is the breakdown of the entire network in the event of a station failure resulting in total shut-down until repairs can be made. This dissertation is a study of designs for fail-safe local area networks. Two methods for maintaining the operation of a ring network are analyzed and compared. The first method uses a double-ring where the inner loop is optically coupled to the outer ring whenever a node failure occurs. The analysis performed here in terms of the scantime reveals that, for a breakdown in the network, the additional delay is at most equal to the single channel scantime delay. When "a" (defined as the propagation time divided by the transmission time) is less than 1, the throughput for the double-ring is comparable to that for the single-ring for all protocols studied. However, for "a" greater than 1, the throughput is lower. For the token-passing ring protocol, the throughput drops to a value of half the single-ring value. The second method uses spokes in the ring that are activated to make a logical ring network connection when a node fails. This design's performance is superior to that of the double-ring in scantime, efficiency under heavy load, and throughput performance. The analysis is based on the assumption that the activation time for the spoked-ring and the double-ring reconfiguration time are negligible. The redundancy in double-ring and spoked-ring networks that allows network operation during station breakdown is not put to any use during normal conditions. This dissertation presents a new design of ring operation where normal operation does use the 'redundant' hardware. A network called dual access bi-ring network (DABNET) is proposed where the stations are divided into two sets (rings) so that most of the communication is ordinarily between stations of each set. A protocol is designed so that communication across rings is allowed. This new concept of two rings that allows communication across them in the event of ring breakdown or when stations across the rings wish to transfer information could be generalized to more than two rings, though the protocols governing such rings would be quite complex.