Tradeoff of horizontal decomposition versus vertical stacking in rearrangeable nonblocking networks

Abstract

It is noted that, when choosing between strictly nonblocking and rearrangeable nonblocking architectures, the following dilemma is often faced: strictly nonblocking networks require too much hardware and the rearrangement occurs too often in rearrangeable nonblocking networks. A class of fault-tolerant switching networks is proposed where the design principle is based on the combination of horizontal decomposition and vertical stacking. This approach makes it possible to make the best tradeoff between the hardware cost and the frequency of rearrangement activities in the network. In addition, the proposed networks are self-routing and do not possess the time and space bottlenecks in conventional switching architectures.<>