Performance Of Developed Model Research Articles

Automatic tuning of sparse matrix-vector multiplication on multicore clusters

To have good performance and scalability, parallel applications should be sophisticatedly optimized to exploit intra-node parallelism and reduce inter-node communication on multicore clusters. This paper investigates the automatic tuning of the sparse matrix-vector (SpMV) multiplication kernel implemented in a partitioned global address space language, which supports a hybrid thread- and process-based communication layer for multicore systems. One-sided communication is used for inter-node data exchange, while intra-node communication uses a mix of process shared memory and multithreading. We develop performance models to facilitate selecting the best configuration of threads and processes hybridization as well as the best communication pattern for SpMV. As a result, our tuned SpMV in the hybrid runtime environment consumes less memory and reduces inter-node communication volume, without damaging the data locality. Experiments are conducted on 12 real sparse matrices. On 16-node Xeon and 8-node Opteron clusters, our tuned SpMV kernel gets on average 1.4X and 1.5X improvement in performance over the well-optimized process-based message-passing implementation, respectively.

Framework for Development and Comprehensive Comparison of Empirical Pavement Performance Models

AbstractEmpirical performance-prediction models are a central part of every network-level pavement management system. In this regard, a variety of novel techniques including computational intelligence have been applied, mainly without a systematic approach to ensure compliance with principles of pavement engineering. In this study, a framework is provided for development and comprehensive comparison of alternative techniques for pavement performance modeling. As an example, several machine-learning techniques are compared in developing flexible pavement-roughness prediction models using Federal Highway Administration (FHWA’s) long-term pavement performance (LTPP) data. Three important principles of model development—maximum likelihood, consistency, and parsimony—are considered in providing a robust parameterization guideline. Variant architectures of artificial neural networks (ANN), radial basis function (RBF) networks, and support vector machines (SVM) are tested to determine the optimum parameters. Fin...

Model and analysis of performance for the method of characteristics direction probabilities with boundary averaging

The work presents a performance model of the method of characteristic direction probabilities (CDP) which integrates the benefits of the collision probability method (CPM) and the method of characteristics (MOC) for solution of the integral form of the Boltzmann Transport Equation and has been implemented in the Michigan PArallel Characteristic based Transport (MPACT) code for 2-D and 3-D transport calculations. The process of boundary averaging reduced the storage and computation but the capability of dealing with complicated geometries is preserved since the same ray tracing information is used as in MOC. The benefits of CDP are demonstrated by the developed performance model which describes the storage, floating point operations and calculation time. The numerical results are given for different cases to show the accuracy, storage, floating point operations and computing time of the CDP compared to the MOC using the performance model. From the cases examined, the boundary average method shows significant improvement on the storage and computational efficiency for three-dimensional cases with sufficient accuracy.

Stream-Based Extreme Learning Machine Approach for Big Data Problems

Big Data problems demand data models with abilities to handle time-varying, massive, and high dimensional data. In this context, Active Learning emerges as an attractive technique for the development of high performance models using few data. The importance of Active Learning for Big Data becomes more evident when labeling cost is high and data is presented to the learner via data streams. This paper presents a novel Active Learning method based on Extreme Learning Machines (ELMs) and Hebbian Learning. Linearization of input data by a large size ELM hidden layer turns our method little sensitive to parameter setting. Overfitting is inherently controlled via the Hebbian Learning crosstalk term. We also demonstrate that a simple convergence test can be used as an effective labeling criterion since it points out to the amount of labels necessary for learning. The proposed method has inherent properties that make it highly attractive to handle Big Data: incremental learning via data streams, elimination of redundant patterns, and learning from a reduced informative training set. Experimental results have shown that our method is competitive with some large-margin Active Learning strategies and also with a linear SVM.

Comparison of Full-Depth Reclamation with Portland Cement and Full-Depth Reclamation with No Stabilizer in Accelerated Loading Test

Full-depth reclamation (FDR) with or without various stabilizers has been successfully used as a rehabilitation strategy in California since 2001. Long-term field monitoring on several FDR projects that used foamed asphalt with portland cement as the stabilizer combined with a comprehensive laboratory study resulted in the preparation of guidelines and specification language for this rehabilitation strategy in 2008. However, the design criteria were essentially empirical, in line with California design procedures for a rehabilitation project of this level. Recently, interest has grown in the use of cement, engineered emulsion, and no-stabilizer FDR strategies in addition to foamed asphalt and in the use of mechanistic design in a greater range of rehabilitation projects. Consequently, the research initiative was extended to a second phase including accelerated load testing on an instrumented test track constructed with these four FDR strategies to gather data for developing performance models that could be included in mechanistic–empirical rehabilitation design procedures. This paper summarizes results of the second set of tests in this accelerated loading study, which compared no-stabilizer and portland cement strategies. The portland cement stabilized section outperformed the unstabilized section in all measured aspects. The most notable observation was in relation to rutting performance; the unstabilized section reached a terminal rut depth of 13 mm after approximately 490,000 equivalent standard axle loads were applied, compared with the cement section, which had a rut depth of only 3.0 mm after more than 43.3 million equivalent standard axle loads. No cracking was observed on either section at the end of testing. Advantages of using portland cement over unstabilized pulverized material are clearly evident from the results.

Optimization-based methodology for the development of wastewater facilities for energy and nutrient recovery

A paradigm shift is currently underway from an attitude that considers wastewater streams as a waste to be treated, to a proactive interest in recovering materials and energy from these streams. This paper is concerned with the development and application of a systematic, model-based methodology for the development of wastewater resource recovery systems that are both economically attractive and sustainable. With the array of available treatment and recovery options growing steadily, a superstructure modeling approach based on rigorous mathematical optimization appears to be a natural approach for tackling these problems. The development of reliable, yet simple, performance and cost models is a key issue with this approach in order to allow for a reliable solution based on global optimization. We argue that commercial wastewater simulators can be used to derive such models, and we illustrate this approach with a simple resource recovery system. The results show that the proposed methodology is computationally tractable, thereby supporting its application as a decision support system for selection of promising resource recovery systems whose development is worth pursuing.

Development of Piecewise Linear Performance Models for Flexible Pavements Using PMS Data

AbstractThis study presents a method to develop piecewise linear (PL) performance models for pavement condition data in a pavement management system (PMS). These condition data are usually ordinal and have more than two severity levels. Ordinal logistic regression is conducted to derive probabilities of each individual severity level. The intersections of probability curves are identified as the breakpoints, which can be used to develop PL models. This proposed method was then applied successfully to develop four PL models, for interstates, U.S. routes, North Carolina routes, and secondary routes, using transverse cracking condition data of flexible pavements in a state DOT’s PMS. Results showed that the PL models reflected actual deterioration trends well and that the proposed method is robust.

An analysis of definition and placement of virtual machines for high performance applications on Clouds

SUMMARYThe efficient deployment of high performance computing applications on Clouds offers many challenges, in particular, for communication‐intensive applications. One strategy to mitigate performance overheads caused by high‐communication latency is to schedule requested virtual machines (VMs) effectively onto physical resources by optimizing VM placement. In current approaches based on Infrastructure as a Service resource provisioning, the task of selecting VM profiles is either left to the user or resolved with policies independent from usage patterns, with potential for performance degradation caused by virtualization overhead. Also, most VM placement heuristics disregard the topology of virtual clusters, resource usage patterns of each application, and competing workloads. In this paper, we study the case of scientific applications in virtual clusters by analyzing how different VM profiles and placements can affect observed performance of a parallel application that uses distributed memory. We propose a description for the characteristics of a virtual cluster through the placement of virtual cores, analyze different configurations with a software for systematic execution of virtual clusters, and explain the observed performance in terms of virtual cluster features and resource usage patterns of the application. Our analysis shows that other factors besides the number of cores can have significant effect on performance, such as virtual core mappings and VM spreading. We discuss how our methodology can be extended toward developing performance models, which are aware of resource contention and virtual cluster topology. Copyright © 2014 John Wiley & Sons, Ltd.

Air Transport Performance and Efficiency: MCDA vs. DEA Approaches

The general aim of this work is the development of airports performance and efficiency predictive models using robust but flexible methodologies, and incorporating traditional indicators as well as new constraints. Particularly it shows and compares the efficiency evolution of the same airport along several years, under several constraints, based on two multidimensional tools: Multi Criteria Decision Analysis (MCDA, by Measuring Attractiveness by a Categorical Based Evaluation Technique - MACBETH) and Data Envelopment Analysis (DEA).The results evidence how MACBETH (MCDA) approach seems to be a very promising one when compared with those (DEA based) traditionally in use.

An analysis of maximum likelihood estimation method for bit synchronization and decoding of GPS L1 C/A signals

In weak GNSS signal environments, extending integration time is paramount to improving the GNSS receiver’s sensitivity. Furthermore, sufficient coherent integration can help to mitigate multipath and cross-correlation false locks and avoid squaring loss. However, extending integration time is limited by the navigation message data bit, if present. The maximum likelihood (ML) estimation method has been shown as the most effective way to estimate the navigation bit boundary locations (i.e., bit synchronization) and subsequently estimate the data bit values (i.e., bit decoding) in the presence of noise alone. In this paper, the performance of ML bit synchronization and decoding is systematically assessed as a function of the number of data bits, the effect of Doppler error and received signal power in different tracking modes (i.e., phase-locked mode and frequency-locked mode). In addition, the theoretical performance models of ML bit synchronization and decoding are developed based on statistical theory. The experimental validation of the developed performance models and analyses is reported. For GPS L1 C/A signals, it is shown that for ML bit synchronization, using 100 data bits, the successful synchronization rate (SSR) can reach to about 100% with C/N0 as low as 20 dB-Hz with no Doppler error. The performance degradation caused by Doppler error is not significant if the Doppler error is within 5 Hz, and with the maximum tolerance of 25 Hz, while for ML bit decoding, the successful decoding rate (SDR) of the 2-bit sequence can reach to about 100% with C/N0 as low as 25 dB-Hz with no Doppler error. The performance degradation caused by Doppler error is not significant if the Doppler error is within 2 Hz. Both theoretical and simulation results establish that the upper bound of Doppler error for a 2-bit sequence is 12.5 Hz.

Comparison of Full-Depth Reclamation with Foamed Asphalt and Full-Depth Reclamation with No Stabilizer in Accelerated Loading Test

Full-depth reclamation (FDR) with foamed asphalt has been successfully used as a rehabilitation strategy in California since 2001. Long-term field monitoring on several projects and a comprehensive laboratory study resulted in the preparation of guidelines and specification language in 2008. However, the design criteria were essentially empirical, in line with California design procedures for this level of rehabilitation project. Recently, there has been growing interest in the use of cement, engineered emulsion, and no-stabilizer full-depth reclamation strategies in addition to foamed asphalt and in the use of mechanistic design in rehabilitation projects. Consequently, the research initiative was extended to a second phase to include accelerated load testing on an instrumented test track constructed with these four FDR strategies to gather data for the development of performance models that can be included in mechanistic–empirical rehabilitation design procedures. This paper summarizes the results of the first two tests in this accelerated loading study, which compared no stabilizer and foamed asphalt–cement strategies. The foamed asphalt section outperformed the unstabilized section in all measured aspects. The most notable observation was in rutting performance: the unstabilized section reached a terminal rut depth of 13 mm after approximately 490,000 equivalent standard-axle loads had been applied, compared with the foamed asphalt and cement section, which had a rut depth of only 4.3 mm after more than 17.7 million equivalent standard-axle loads. No cracking was observed on either section. The advantages of using foamed asphalt with cement over unstabilized pulverized material are clearly evident from the results.

Black box scheduling for resource intensive virtual machine workloads with interference models

Modern datacenters consist of increasingly powerful hardware. Achieving high levels of utilization on this hardware often requires the execution of multiple concurrent workloads. Virtualization has emerged as an efficient means to isolate workloads by partitioning large physical resources using self-contained virtual machine images. Despite the many advantages, some challenges regarding performance isolation still need to be addressed. Unmanaged multiplexing of resource intensive workloads has the potential to cause unexpected variances in workload performance.In this paper, we address this issue using performance models based on the runtime characteristics of virtualized workloads. A set of resource intensive workloads is benchmarked with increasing degrees of multiplexing. Resource usage profiles are constructed using the metrics made available by the Xen hypervisor. Based on these profiles, performance degradation is predicted using several existing modeling techniques. In addition, we propose a novel approach using both the classification and regression capabilities of support vector machines. Application clustering is used to identify several application types with distinct performance profiles. Finally, we evaluate the developed performance models by introducing several new scheduling techniques. We demonstrate that the integration of these models in the scheduling logic can significantly improve the overall performance of multiplexed workloads.

Data-driven models of steady state and transient operations of spiral-wound RO plant

The development of data-driven RO plant performance models was demonstrated using the support vector regression model building approach. Models of both steady state and unsteady state plant operation were developed based on a wide range of operational data obtained from a fully automated small spiral-wound RO pilot. Single output variable steady state plant models for flow rates and conductivities of the permeate and retentate streams were of high accuracy, with average absolute relative errors (AARE) of 0.70%–2.46%. Performance of a composite support vector regression (SVR) based model (for both streams) for flow rates and conductivities was of comparable accuracy to the single output variable models (AARE of 0.71%–2.54%). The temporal change in conductivity, as a result of transient system operation (induced by perturbation of either system pressure or flow rate), was described by SVR model, which utilizes a time forecasting approach, with performance level of less than 1% AARE for forecasting periods of 2s to 3.5min. The high level of performance obtained with the present modeling approach suggests that short-term performance forecasting models that are based on plant data, could be useful for advanced RO plant control algorithms, fault tolerant control and process optimization.

Potential of SOFC CHP systems for energy-efficient commercial buildings

An energy-efficient and low-emissions solid oxide fuel cell (SOFC) combined heat and power (CHP) system is a promising electric and thermal energy generation technology for implementation in future commercial buildings.Recently developed performance models for kW- and MW-sized SOFC CHP systems have been integrated into a building energy simulation model of a medium-sized (7000m2) office building to evaluate the potential of the system to lower annual utility costs and to reduce CO2 emissions. An optimized 175kW SOFC CHP system successfully lowered annual utility costs by up to 14.5% over a baseline HVAC system in locations with space heating-dominant loads. Potential CO2 emissions were reduced by up to 62% over the baseline case for the optimized SOFC CHP systems.If the capital and installation costs of SOFC CHP systems are reduced in the near future, through the realization of cheaper, high-performance intermediate temperature SOFCs, the SOFC CHP system may be a promising energy-efficient and low-emitting alternative power and thermal energy cogeneration technology.

Development of decay performance models for outdoor timber structures

Performance based building (PBB) and design is closely connected to various needs and requirements: Performance levels need to be defined, test methods for verification of performance need to be developed, and reliable performance data are needed for materials, products, constructions, and different design solutions. In contrast to other building materials, PBB and thus service life prediction of timber and wood-based products requires particular consideration of wood-degrading organisms and their physiological needs. For the most relevant group of wood-destroying organisms, which are the different decay fungi and bacteria, wood moisture content and temperature need to be considered as key factors. This study aimed on the development of performance models on the basis of hard data obtained in field trials performed under most realistic conditions. Dose–response relationships, which can serve as essential parts of a performance based design model, were derived from material climatic data and corresponding decay development in the field. Different dose–response models are proposed and evaluated for predicting onset (and progress) of decay when wood is exposed to a dynamic and arbitrary climate exposure described in terms of time series of coupled temperature and moisture content. A logistic dose–response model was primarily focused on describing the relation between exposure and decay rating for moisture traps with long periods of high moisture contents. A two-step linear engineering model was more focused on predicting the behavior in a wider, more simplified, sense where periods of high moisture content are interrupted by periods of drier and/or colder climate.

Pervious concrete pavement performance modeling: an empirical approach in cold climates

Pervious concrete pavement (PCP) is an appropriate means to meet growing environmental demands. To apply pervious concrete as a pavement, its performance should be studied over its life span. This research aims to develop empirical performance models based on a proposed condition index through incorporation of integrated laboratory and field work. Since no condition index has been developed for PCP to date, first, a condition index is developed. The condition index is proposed as a combination of two indices: a surface distress index and a functional performance index. Two sources of data have been collected to develop performance models including panel rating and field investigations. Performance models are developed in two phases using regression analysis techniques. In phase I, performance models are presented as functions of a surface distress index and a functional performance index, while in phase II, performance models are correlated with pavement age and successfully validated.

DEVELOPMENT OF PERFORMANCE MODELS FOR A TYPICAL FLEXIBLE ROAD PAVEMENT IN NIGERIA

The results of a study conducted to facilitate the development of road pavement performance models that are appropriate for Nigeria and similar developing countries and could predict the rate of deterioration over their lifespan have been presented. Comprehensive investigations were carried out on the expressway linking Lagos (the economic nerve centre of Nigeria) with Ibadan (the largest city in West Africa) - apparently one of the most heavily trafficked roads in the country. Data relating to traffic characteristics, pavement condition ratings, distress types, pavement thickness, roughness index, rainfall and temperature, were collected. Models were developed to determine Pavement Condition Score (PCS) and International Roughness Index (IRI). Stepwise Regression was used to analyse the data and quantify the impact of key input parameters on the PCS and IRI. Parameters such as depth of ruts and area of pot holes were found to be statistically significant in predicting PCS while number of patches, length of longitudinal cracks and depth of ruts were statistically significant in predicting IRI. The models can be used for planning road maintenance programs, thus minimizing the need for comprehensive data collection on pavement condition before the maintenance exercise, which is costly and time consuming.

Performance modelling of anonymity protocols

Anonymous network communication protocols provide privacy for Internet-based communication. In this paper, we focus on the performance and scalability of anonymity protocols. In particular, we develop performance models for two anonymity protocols from the prior literature (Buses and Taxis), as well as our own newly proposed protocol (Motorcycles). Using a combination of experimental implementation, simulation, and analysis, we show that: (1) the message latency of the Buses protocol is O(N2), scaling quadratically with the number of participants; (2) the message latency of the Taxis protocol is O(N), scaling linearly with the number of participants; (3) the message latency of the Motorcycles protocol is O(log2N), scaling logarithmically with the number of participants. Motorcycles can provide scalable anonymous network communication, without compromising the strength of anonymity provided by Buses or Taxis.

Analytical models for understanding space, backoff, and flow correlation in CSMA wireless networks

In wireless networks employing carrier-sense multiple-access with collision avoidance (CSMA/CA), correlations between the service processes of different nodes arise as a result of competition for common wireless channels and the dependencies between upstream and downstream traffic flows. These dependencies make the development of tractable performance models extremely difficult. To address this purpose, we present a new continuous-time model for CSMA wireless networks where we combine a node model and a channel model in order to capture correlation. Simplification methods are presented that make our models computationally tractable for large networks with minimal loss of accuracy. The model can be used for both single and multi-hop wireless networks and takes into account non-saturated queues, backoff-stage dependence of collision probabilities, and the correlation between departure processes and arrival processes of adjacent nodes. The model can be used to compute probabilistic quality of service guarantees to optimize end-to-end throughput and end-to-end delay by adjusting arrival and backoff rates along various paths.

Performance comparison of parallel eigensolvers based on a contour integral method and a Lanczos method

Abstract We study the performance of a parallel nonlinear eigensolver SSEig which is based on a contour integral method. We focus on symmetric generalized eigenvalue problems (GEPs) of computing interior eigenvalues. We chose to focus on GEPs because we can then compare the performance of SSEig with that of a publicly-available software package TRLan, which is based on a thick restart Lanczos method. To solve this type of problems, SSEig requires the solution of independent linear systems with different shifts, while TRLan solves a sequence of linear systems with a single shift. Therefore, while SSEig typically has a computational cost greater than that of TRLan, it also has greater parallel scalability. To compare the performance of these two solvers, in this paper, we develop performance models and present numerical results of solving large-scale eigenvalue problems arising from simulations of modeling accelerator cavities. In particular, we identify the crossover point, where SSEig becomes faster than TRLan. The parallel performance of SSEig solving nonlinear eigenvalue problems is also studied.