Mehrnaz Moudi

x-Folded TM

Massively parallel computers (MPCs) are currently being actively studied. Interconnection networks are used for the connection of a significant number of processors in such parallel systems because they are introduced as one of the key elements in parallel processing. Meshes, torus, and hypercubes were initially introduced as simple topologies of interconnection networks. At present, these have been replaced by more complicated topologies that also exhibit high performance. Because the diameter, average distance, and cost of topologies have a marked influence on network performance, this paper presents a novel topology called x-Folded TM, which is a TM topology that is folded according to the x-axis. For an n × n network, the diameter of an x-Folded TM is less than that of a TM and a torus, and the average distance of an x-Folded TM is less than that of a mesh, torus, and TM. Compared with mesh, torus, and TM networks, a Folded TM network presents reductions in the average distance, diameter, and cost, which accounts for its efficient performance. As observed from the presented simulation results, the performance of an x-Folded TM is similar to that of most torus networks and better than that of some torus networks. The results verify the effectiveness of the x-Folded TM network, as determined through its comparison with networks with other topologies.

A Cost-Effective Architecture For Optical Multistage Interconnection Network

In this paper a new architecture for Optical Multistage Interconnection Networks (OMINs) has been proposed to avoid crosstalk problem. At the same time, the probablity of losing pass through an optical long connection path is reduced in this architecture. The new architecture is inherent form the standard OMIN by converting two switches of the network to one switch in each row. By reducing the number of switches in new architecture, the reduction in the execution time is considered. The modifying in the number of passes via the same low stage transformation is negligible. The ability of the new architecture to decrease cost and avoid crosstalk has been validated through simulations that show improvement in the network performance in terms of approximately 30% reduction in the execution time.

A GreedyZero algorithm to minimise the conflicts in an Optical Multistage Interconnection Network

An Optical Multistage Interconnection Network (OMIN) is an important class of Interconnection Network that has large transmission capacity in the communication networks. One of the severe problems is the conflict which is caused through coupling two signals within switching elements. In this paper, we have proposed a new algorithm to minimise the number of conflicts of OMINs using the greedy graph colouring and Zero algorithm. The greedy graph colouring approach employed the Zero algorithm to categorise inputs of the OMINs without conflict. The results showed that the number of passes reduced by approximately 30%. In addition, the average execution time of our proposed algorithm was less than the average execution time of the Zero algorithm.

Adaptive routing algorithm in x-Folded TM topology

In recent days, more complicated interconnection topologies have been replaced with previously simple topologies that also exhibit high performance. x-Folded TM network is a TM network that is folded according to the imaginary x-axis in interconnection networks. For a x-Folded TM network, there is a reduction in the average distance and diameter which corroborates the efficient performance. Also many routing algorithms can be applied to interconnection network for the efficient use of network resources. This paper presents the effectiveness of the x-Folded TM network with the average latency and network throughput under the adaptive routing algorithm to improve dynamic communication performance. We evaluate the communication performance with simulating the topology under the adaptive routing algorithm. It is found from the result that the performance of a x-Folded TM network has been improved compared with other topologies.

Mathematical Modelling of Wormhole-Routed x-Folded TM Topology in the Presence of Uniform Traffic

Recently, x-Folded TM topology was introduced as a desirable design in k-ary n-cube networks due to the low diameter and short average distance. In this article, we propose a mathematical model to predict the average network delay for (\(k \times k\)) x-Folded TM in the presence of uniform traffic pattern. Our model accurately formulates the applied traffic pattern over network virtual channels based on the average distance and number of nodes. The mathematical results indicate that the average network delay for x-Folded TM topology is reduced when compared with other topologies in the presence of uniform traffic pattern. Finally, the results obtained from simulation experiments confirm that the mathematical model exhibits a significant degree of accuracy for x-Folded TM topology under the traffic pattern even in varied virtual channels.

A Performance Comparison of Deterministic and Adaptive Routing for an x-Folded TM Topology

This article focuses on a novel topology in interconnection networks which was called x-Folded TM. We explored the dynamic communication performance of this topology under various traffic patterns by designing new scheme for routing algorithms. In this scheme, the topology is partitioned to make the routing without deadlock in x-Folded TM topology. Using deterministic and adaptive routing, we have evaluated the performance of x-Folded TM topology in compare with Torus. A comparison of results reveals that the average delay grows steadily with the increase of injection rate until reaching saturation, which is the smallest rate of traffic for channel saturation in the network. Therefore, the achievements introduce x-Folded TM topology as efficient one in interconnection networks.

Mathematical modelling for TM topology under uniform and hotspot traffic patterns

ABSTRACT Interconnection networks are introduced when dealing with the connection of a significant number of processors in massively parallel systems. TM topology is one of the latest interconnection networks to solve the deadlock problem and achieve high performance in massively parallel systems. This topology is derived from a Torus topology with removing cyclic channel dependencies. In this paper, we derive a mathematical model for TM topology under uniform and hotspot traffic patterns to compute the average delay. The average delay is formulated from the sum of the average delay of network, the average waiting time of the source node and the average degree of virtual channels. The results obtained from the mathematical model exhibit a close agreement with those predicted by simulation. In addition, sufficient simulation results are presented to revisit the TM topology performance under various traffic patterns.

Analyzing the performance of low stage interconnection network

In order to avoid crosstalk, a new architecture is proposed for Optical Multistage Interconnection Networks (OMINs). In the new architecture, two switches are replaced by one switch in each row. Reduction in the number of switches makes the considerable reduction in the execution time. To study the performance of the new architecture, analytical techniques also can be used effectively. The theory of probability is used to derive mathematical equation for network bandwidth allocation of a unit load. The obtained results show the improvement in the network performance. By increasing load, the bandwidth is reduced. In addition the simulation is applied to validate the new architecture and show improvement in the performance by approximately 30 % reduction in the execution time.