Performance Evaluation Of Interconnection Networks Research Articles

The modeling power of CINSim: Performance evaluation of interconnection networks

CINSim is a simulator for modeling a great variety of interconnection networks that are composed of switches, buffers, sources, and destinations. This includes regular networks as well as irregular ones. Steady-state simulation can be applied as well as terminating simulation to observe transient behavior of interconnection networks. Due to the involved stochastic events, e.g. randomly generated traffic, confidence levels and estimated precisions are calculated to determine the accuracy of the results. Additionally, quantiles are provided allowing to estimate the distribution of results. Dynamic reconfiguration of interconnection networks can also be modeled by this simulator. Thus, the reconfiguration process can be observed to investigate the particular network behavior during this time interval. CINSim can be executed on parallel or distributed systems to speed up simulations. To the best of the authors’ knowledge, CINSim is the first simulator that combines those features with a convenient graphical user interface.

Toward a more realistic performance evaluation of interconnection networks

lnterconnection network design plays a central role in the design of parallel systems. Most of the previous research has evaluated the performance of interconnection networks in isolation. In this study, we investigate the relationship between application program characteristics and interconnection network performance using an execution driven simulation test bed: the Reconfigurable Architecture Workbench (RAW). We simulate five topological configurations of a k-ary n-cube interconnect and four different network link models for a 4,096 node SIMD machine, and quantify the impact of the network on two application programs. We provide experimental evidence that such in-context simulation provides a better view of the impact of network design variables on system performance. We show that recent results, indicating that low-dimensional designs provide better ICN performance, ignore application requirements that may favor high-dimensional designs. Furthermore, applications that would appear to favor low dimensional designs may not, in fact, be significantly impacted by the network's dimensionality. We experimentally test the results of published performance models comparing the use of a synthetic load to that of a load generated by a typical application program.

Probability distributions of interdeparture time and response time in multipacket CSMA/CD systems

An exact analysis is presented of an asynchronous (unslotted) multipacket CSMA/CD-DFT (carrier-sense multiple-access/collision detection with delayed first transmission) model and used to derive (1) the Laplace-Stieltjes transform of the probability distribution function (PDF) of the packet interdeparture time, (2) the LST of the PDF of the message interdeparture time, (3) the moment generating function (MGF) of the message response time, and (4) the MGF of the message response time conditioned on the number of packets necessary for transmission of a message. Higher moments as well as averages of these performance measures are important for performance evaluation of interconnected networks or integrated networks of voice and data. As the performance model allows each message to consist of a geometrically distributed number of constant-length packets, the result is also useful in examining the performance of error-prone CSMA/CD systems as well as the impact of random message sizes on the system performance. Numerical examples include comparisons between the DFT model and the IFT (immediate-first-transmission) model.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Analysis of interconnection networks with different arbiter designs

In performance evaluation of interconnection networks, it is usually assumed that in the event of conflicts a request is accepted with equal probability. This paper illustrates that this arbitration policy discriminates against remote or less frequent requests because it rejects them most of the time. A favorite memory environment is considered as an example and the network performance under various arbitration policies is examined. Policies of equal acceptance and priority to favorite and nonfavorite requests are examined by defining three probabilities of acceptance. The networks considered are crossbar, multiple-bus, and multistage interconnection networks. For these networks arbiter designs that incorporate the priorities are presented.