Please choose your delivery country and your customer group
The purpose of this study was to examine the use of distributed adaptive routing algorithms on concurrent class computers. The implemented routing algorithm allowed each node to select the next node based on two criteria: the fewest number of hops; and the smallest delay time. This study was limited to the comparison of a distributed adaptive routing algorithm, implemented at the applications layer, with the current static routing and with a simulation of the current routing implemented at the applications layer. The comparison with the simulated current static routing provides a measure of the possible performance gain had the adaptive routing algorithm been implemented at the network layer. Each of three configuration was comprised of four processes: a Host Process, a Routing Process, a Ring Control Process, and a Network Loading Process. The Host Process controlled the loading of the processes onto the iPSC, the Routing Process controlled the message routing, the Ring Control Process provided the baseline message passing, while the Network Loading Process provided communications congestion on selected links. The metric used to compare the Routing Process performance was the average delay time for passing a message around the ring.