The shuffle/exchange-plus networks

Abstract

From the r e l i a b i l i t y point of view, the uniqueness of path between any processor and memory module in a standard k-column Shuffle/Exchange (S/E) network is an inherent weakness. It is proposed to add a column of switches to the existing multistage S/E network such that the modified network, which will be called the S/E-Plus network, will retain the permuting power of the corresponding S/E network and have dual paths between any processor and any memory module. Fault tolerant routing algorithm is then designed to exploit the dual path structure of the S/E-Plus network. Various applications of the routing algorithm including a technique for performing permutation in the S/E-Plus network in which a fault exists are also discussed. Finally, the path r e l i a b i l i t y and the path r e l i a b i l i t y gain are defined to evaluate the S/E-Plus network. P e r m i s s i o n to copy w i t h o u t f e e a l l or p a r t o f t h i s m a t e r i a l i s granted provided t h a t t h e c o p i e s are not made or d i s t r i b u t e d f o r d i r e c t commercial a d v a n t a g e , the ACH c o p y r i g h t

Multistage Interconnection Networks Reliability Evaluation Based on Stratified Sampling Monte Carlo Method

“…For terminal reliability, it is compared against terminal reliability of SEN+ calculated using the mathematical approach [11,12].…”

“…The exact terminal reliability for SEN+ can be calculated by simplifying the two path connections between the source and destination terminals into two parallel series of SE connections [11,12] as shown in Fig. 4.…”

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

“…The extra stage of the ESC does increase the likelihood of a fault compared to the Generalized Cube due to the additional hardware. However, analysis of an independently developed related network shows that for reasonable values of interchange box reliability there is a gain in network reliability as a result of an extra stage [27]. It should also be noted that a failure in a stage n multiplexer or stage 0 demultiplexer has the effect of a link fault, which the ESC can tolerate as shown in this section.…”

Performance and fault tolerance improvements in the Inverse Augmented Data Manipulator network

“…Fault detection and fault tolerance in interconnection networks have been investigated by a number of researchers [1,7,[13][14][15][16]18,21,25] and surveyed by [2].…”

“…(2) Some papers discuss multistage, interchange box (B-element [18]) based cube/Benes type networks [1,18,25]. Most of these networks have a sufficient bandwidth but fault tolerance is achieved by passing through an extra stage of interchange boxes [1,25] or by making multiple passes through the network [18].…”