Rosa Alcover

A New Reliability Model for Interconnection Networks

The traditional approach to study fault-tolerance in multicomputer interconnection networks consists of determining the worst possible combination of faulty components that causes a network failure, and then assuming that this will occur. But the worst possible combination does not always occur, and the routing algorithm allows the network to work in the presence of a greater number of failures. Thus network reliability parameters computed according to the traditional approach will be underestimated. In this paper we propose a new methodology to compute accurately the reliability function. The reliability parameters have been computed for an interconnection network with mesh topology, taking into account size, routing algorithm, failure and repair rates of the network channels and coverage.

Optimizing network throughput: optimal versus robust design

Interconnection network performance is usually measured in terms of its latency (time required to deliver a message) and throughput (maximum traffic accepted by the network). At first glance, minimizing average message latency is the main designer goal, because average network traffic is usually far from saturation. However, applications can also generate very high peak traffic. In order to deal with such situations, it is important that network throughput is also high. On the other hand, interconnection network performance depends on several parameters. Some of them can be chosen by the designer: routing algorithm, switching technique, topology and node design parameters. However, there are other parameters that cannot be selected by the designer. Among these, there are parameters that depend on the application, such as message size, message destination distribution and message traffic, as well as parameters defined by the customer, such as network size. Network designer can select the design parameters that maximize average (optimal design) or the design parameters that achieve a good performance under all the feasible combinations of the parameters that cannot be selected by him (robust design). Notice that both alternatives do not always lead to the same parameter configuration. Previously we chose the design parameters of a k-ary n-cube network considering optimize latency. In this case, optimal and robust design lead to the same choice. In this paper, we obtain these design parameters considering optimized network throughput. Unfortunately, there is a discrepancy between optimal and robust design criteria, being the former the best choice.

A cost-effective methodology for the evaluation of interconnection networks

Abstract Interconnection network performance depends on several parameters, including network design parameters, network size, message traffic and message length. Simulation is the methodology usually followed in evaluation studies, because the model can more faithfully represent hardware implementation, taking into account more details. Nevertheless, the number of parameter combinations is often very high, and simulations also take long to complete. Therefore, evaluation studies must choose a subset of the parameters and restrict the variability of each of them. In this paper, we propose a methodology for evaluating interconnection networks. It is based on experimental design used in statistical studies. Using this methodology, we can study network behavior considering many parameters, running only a subset of the simulations required to study all the combinations. In addition, the methodology permits to quantify the effect of interactions among the parameters. We apply this methodology to adjust node design parameters such as number of virtual channels, input buffer size, and output buffer size for a 8-ary 3-cube with adaptive (both partially and fully) wormhole routing. We show that running only one third of the simulations required to study all the combinations, the most significant effects can be estimated without a noticeable loss in precision.

Dependability Analysis of a Fault-Tolerant Network Reconfiguring Strategy

Fault tolerance mechanisms become indispensable as the number of processors increases in large systems. Measuring the effectiveness of such mechanisms before its implementation becomes mandatory. Research toward understanding the effects of different network parameters on the dependability parameters, like mean time to network failure or availability, becomes necessary. In this paper we analyse in detail such effects with a methodology proposed previously by us. This methodology is based on Markov chains and Analysis of Variance techniques. As a case study we analyse the effects of network size, mean time to node failure, mean time to node repair, mean time to network repair and coverage of the failure when using a 2D mesh network with a fault-tolerant mechanism (similar to the one used in the BlueGene/L system), that is able to remove rows and/or columns in the presence of failures.

Accurate reliability and availability models for direct interconnection networks

Fault tolerance in multicomputer interconnection networks has been traditionally studied by determining the worst possible combination of faulty components that causes its failure and then assuming that this will occur. But, the probability of the worst possible combination is usually low, and the routing algorithm may be able to find a route between source and destination nodes. The network dependability parameters computed according to this approach will be underestimated. In this paper we propose a methodology for accurately evaluating interconnection network dependability. In addition, we apply it to obtain an accurate estimation of the reliability and availability parameters in a 2-D mesh, taking into account network size, routing algorithm, failure and repair rates of nodes, and coverage. Finally we compare the computed results under both approaches.

Interconnection network design: a statistical analysis of interactions between factors

Interconnection network performance depends on several parameters, including network design parameters, network size, message traffic and message length. Simulation is the methodology usually followed in evaluation studies, because the model can more faithfully represent hardware implementation, taking into account more details. Nevertheless, the number of parameter combinations is often high, and simulations also take long to complete. Therefore, evaluation studies must choose a subset of the parameters and restrict the variability of each of them. In a previous paper (IEEE Computer Soc. TCCA Newsletter, pp. 32-37, Aug. 1995), we have proposed a methodology for evaluating interconnection networks. It is based on experimental design used in statistical studies. Using this methodology, we can study network behavior considering many parameters, running only a subset of the simulations required to study all the combination. In addition, the methodology permits us to quantify the effect of interactions among the parameters. In this paper, we make use of the second advantage of this methodology, analysing the effect of node design parameters and their interactions for an 8-ary 3-cube with adaptive wormhole routing.

Support planning to the implementation of School of Engineering in Computer Science - An oriented approach to teacher training

Main objective of Action Plan for European Convergence (PACE) that Universitat Politecnica de Valencia (UPV) has been developing since academic year 2005-2006 has been to facilitate the process of adaptation of new courses of bachelors degree to the European Higher Education Area (EEES). To carry out this ambitious project, centers have identified a set of coordinated actions by teams of teachers interested in teaching innovation. One of the activities promoted by School of Engineering in Computer Science (ETSINF) of UPV is a training plan for teachers. The main objective is to provide teachers with tools that facilitate the design of new courses of Bachelors Degree in Computer Engineering under the current regulations, with particular emphasis on two aspects: the use of active methodologies and the evaluation criteria. This article describes the training activities that have taken place in the ETSINF during academic year 2009-2010 and the results of opinion poll for teachers about the appropriateness of these actions and their impact on the process to adapt their courses.

Improving the Accuracy of Reliability Models for Direct Interconnection Networks

Fault-tolerance in multicomputer interconnection networks has been traditionally studied by determining the worst possible combination of faulty components that causes a network failure and then assuming that this will occur. But, the worst possible combination may occur with low probability and the routing algorithm may allow the network to work, even when there is a large number of faults. Thus, the network dependability parameters computed according to this approach will be underestimated. In a previous paper [3], we have proposed a new methodology, based on Markov chains, for evaluating interconnection network dependability. Using this methodology, we can accurately compute the network reliability behavior. In this paper we apply it to evaluate dependability parameters in a 2-D mesh, taking into account network size, routing algorithm, failure and repair rates of nodes and coverage. Finally, we compare the computed results under traditional and our approach.

An accurate analysis of reliability parameters in meshes with fault-tolerant adaptive routing

The traditional approach to study fault-tolerance in multicomputer interconnection networks consists of determining the worst possible combination of faulty components that causes a network failure, and then assuming that this will occur. But the worst possible combination does not always occur, and the routing algorithm allows the network to work in the presence of a greater number of failures. The network reliability parameters computed according to the traditional approach will be under-estimated. In this paper we use a new methodology to compute accurately the reliability and availability functions. The reliability parameters have been computed for a network with mesh topology, taking into account size, routing algorithm, failure and repair rates of the network channels and coverage.