The Extra Stage Cube: A Fault-Tolerant Interconnection Network for Supersystems

Abstract

Abstract-The Extra Stage Cube (ESC) interconnection network, a fault-tolerant structure, is proposed for use in large-scale parallel and distributed supercomputer systems. It has all of the interconnecting capabilities of the multistage cube-type networks that have been proposed for many supersystems. The ESC is derived from the Generalized Cube network by the addition of one stage of interchange boxes and a bypass capability for two stages. It is shown that the ESC provides fault tolerance for any single failure. Further, the network can be controlled even when it has a failure, using a simple modification of a routing tag scheme proposed for the Generalized Cube. Both one-to-one and broadcast connections under routing tag control are performable by the faulted ESC. The ability of the ESC to operate with multiple faults is examined. The ways in which the ESC can be partitioned and permute data are described.

Tolerating multiple faults in multistage interconnection networks with minimal extra stages

“…None of the previously proposed MINs meets this lower bound for any given f. For example, the ESC solution only works for f I [2]; the number of switch faults that the Error-Correcting Code solutions tolerate is in general less than the redundant stages required [5].…”

“…The faulttolerant capabilities of MIN have been widely studied. In 1982, Adams and Siegel introduced the Extra Stage Cube [2], a construction that tolerates one switch failure with one additional switching stage. In this paper, we study how to construct Multistage Interconnection Networks to tolerate multiple switch faults with extra stages.…”

“…This problem of tolerating a single switch fault with extra stages has been investigated extensively in the past. Adams and Siegel first proposed a solution, called Extra Stage Cube (ESC) [2]. Xu and Shen proposed a solution, which can be viewed as adding an extra stage with an all-1 mask [16].…”

Fast arbitration in dilated routers

“…By randomly choosing the path through these extra switch stages, some of the properties associated with dilation can be achieved, primarily a degree of fault tolerance and hot spot avoidance [6]. The difficulty with such extra stage approaches is that the extra stage route is usually chosen prior to message routing, providing no opportunity to avoid busy ports in the switch components.…”

Performance Analysis of Multilayered Multipriority Asymmetric-Sized Delta Networks

“…MINs are distinguished into two classes: the first class has the Banyan property [2] with its most prominent representatives being delta networks [3], omega networks [4], and generalized cube networks [5]; the second category includes MINs not having the Banyan property, such as augmented and CLOS MINs. Among the two classes, the first one is more widely used, since nonBanyan MINs are generally more expensive and complex.…”