Long-run performance analysis of a multi-scale TCP traffic model

Reconfigurable computing (RC) is rapidly emerging as a promising technology for the future of high-performance and embedded computing, enabling systems with the computational density and power of custom-logic hardware and the versatility of software-driven hardware in an optimal mix. Novel methods for rapid virtual prototyping, performance prediction, and evaluation are of critical importance in the engineering of complex reconfigurable systems and applications. These techniques can yield insightful tradeoff analyses while saving valuable time and resources for researchers and engineers alike. The research described herein provides a methodology for mapping arbitrary applications to targeted reconfigurable platforms in a simulation environment called RCSE. By splitting the process into two domains, the application and simulation domains, characterization of each element can occur independently and in parallel, leading to fast and accurate performance prediction results for large and complex systems. This article presents the design of a novel framework for system-level simulative performance prediction of RC systems and applications. The article also presents a set of case studies analyzing two applications, Hyperspectral Imaging (HSI) and Molecular Dynamics (MD), across three disparate RC platforms within the simulation framework. The validation results using each of these applications and systems show that our framework can quickly obtain performance prediction results with reasonable accuracy on a variety of platforms. Finally, a set of simulative case studies are presented to illustrate the various capabilities of the framework to quickly obtain a wide range of performance prediction results and power consumption estimates.

This special section of Multiscale Modeling and Simulation: A SIAM Interdisciplinary Journal, on multiscale modeling in life and materials sciences, is based on a summer program with the same theme, held August 4--30, 2003, at the Universita della Svizzera Italiana in Lugano, Switzerland, and organized by the ETH Zurich Computational Laboratory (CoLab). This workshop emphasized that multiscale modeling and simulation can serve as a commonscientific language enabling cross-fertilization across disciplines and the definition of new scientific frontiers. Materials science was one of the first disciplines that emphasized the need for novel multiscaling approaches bridging continuum and atomistic models. In life sciences, the rapid growth of the role played by computational methods has led to the early realization that a multiscaling approach is necessary to tackle inherently complex phenomena involving cells, organisms, and ecosystems. This summer program provided the opportunity to scientists from these different scientific fields to exchange ideas and to pursue research in an interdisciplinary fashion using multiscale computational methods as a common language. These interactions demonstrated that research in multiscaling is a fertile breeding ground for the development of novel powerful computational tools for domains where computing is established as a tool for scientific inquiry and for the dissemination of this know-how to disciplines for which computation is an emerging form of scientific investigation. This special section of MMS is a representative sample of theoretical and computational approaches to multiscale modeling and simulation in materials and life sciences. The papers discuss multiscale coupling of physical models and time/space multiresolution techniques. The problem of coupling models across space and time scales is a fundamental issue in multiscale modeling and simulation. Multiscale approaches often require the rigorous integration of diverse physical models, such as the case of blood flow in the circulatory system (Fernandez, Milisic}, and Quarteroni). The multiscale integration of diverse physical models in nanoscale magnetic systems involves the proper coupling of quantum effects in classical descriptions (Wessel, Trebst, and Troyer). Multiscale simulations of materials with heterogeneous properties are tackled using asymptotic expansions for the case of conducting inclusions embedded in a homogeneous matrix phase (Ben Hassen and Bonnetier). A broad class of multiscale problems involves the description of macroscale systems using only the information available by suitable microscopic models. This technique is demonstrated in simulating the evolution of parabolic homogenization problems with nonlinear reactions, using as a microscopic model a partial differential equation with rapidly oscillating coefficients (Samaey, Roose, and Kevrekidis). A related work extends this concept to problems with a wide separation in time scales. Using as an application a model of bacterial chemotaxis it is shown that projective time integration of coarse variables can be carried out in time scales that are long compared to those of the microscopic dynamics (Setayeshgar, Gear, Othmer, and Kevrekidis). In problems with time-dependent, large space scale variations a new class of multiresolution particle methods enhances their inherent adaptivity and presents an alternative to grid-based methods (Bergdorf, Cottet, and Koumoutsakos). Future research in multiscale modeling would continue to rely on synergies across disciplines for the development of novel computational methods, the integration of physical models, and the formulation of new challenges. This principle will continue to guide future workshops in multiscale modeling and simulations at the CoLab and, we believe, an ever-increasing number of scientific institutions. The editors in charge would like to credit, for their hard work and dedication, those editors who helped handle the peer-review process for papers submitted to this special section. Thanks to Emily Carter, Wolfgang Hackbusch, Robert Kohn, Hans Christian Ottinger, Olivier Pironneau, and John Willis. We also wish to thank SIAM Publications Manager Mitch Chernoff for his invaluable advice and help in preparing this section.

For the centrifugal compressor aerodynamic design of a turbocharger, first of all, the works for system matching to the engine specification must be preceded. Then, mean line design together with performance prediction should be carried out for preliminary design. In the mean line prediction, a slip factor is adopted as a function of flow coefficient and geometry instead of Wiesner’s equation, and it is found that the predicted result of slip magnitude is more accurate than that of conventional slip factor. Also, three-dimensional blade profile shape is generated on the basis of the preliminary design. The Navier-Stokes Equation solver with a turbulent model is used to find whether three-dimensionally designed geometry is reasonable by analyzing loading distribution of the blade. By investigating diffuser flow field of the simulated result, the diffuser inlet and exit angles were modified for the flow to move smoothly along the diffuser geometry. Modified performance prediction results shows better than those of original specification. Consequently, off design performance prediction results and numerical simulation result show good agreement with the experimental data. The modified design results show more increased compression ratio and efficiency than those of previous design results. The increased choke margin has made a stable operating range larger.

This work applies a 3D multi-scale bottom-up approach for modeling the processes of diffusion and reaction-diffusion in porous catalyst layers. The performance of the random pore model to predict effective transport coefficients are compared with the results of the multi-scale diffusion model. The results of the 3D multi-scale diffusion model are employed in a 1D pseudo-homogeneous reaction-diffusion model with a relative good agreement with the 3D multi-scale reaction-diffusion model. Furthermore, the former multi-scale model was coupled to a full-scale reactor model with good results and high advantages in terms of computational time savings.

With the development of data mining technology, the analysis of event-related potential (ERP) data has evolved from statistical analysis of time-domain features to data-driven techniques based on supervised and unsupervised learning. However, there are still many challenges in understanding the relationship between ERP components and the representation of familiar and unfamiliar faces. To address this, this paper proposes a model based on Dynamic Multi-Scale Convolution for group recognition of familiar and unfamiliar faces. This approach uses generated weight masks for cross-subject familiar/unfamiliar face recognition using a multi-scale model. The model employs a variable-length filter generator to dynamically determine the optimal filter length for time-series samples, thereby capturing features at different time scales. Comparative experiments are conducted to evaluate the model’s performance against SOTA models. The results demonstrate that our model achieves impressive outcomes, with a balanced accuracy rate of 93.20% and an F1 score of 88.54%, outperforming the methods used for comparison. The ERP data extracted from different time regions in the model can also provide data-driven technical support for research based on the representation of different ERP components.

Lithium-ion batteries are a key enabler for a low-carbon future, however their performance and lifetime are influenced by complex and coupled electrochemical, thermal and mechanical factors across different length and time scales. In this talk, we explore the key degradation modes which limit battery lifetime and how multi-scale models can help describe these effects towards improved cell designs and device control. At the particle level, we explore how phase field fatigue models of cathode particles can be used to understand crack growth, leading to loss of active material [1]. Here, non-linear crack growth is observed due to the fatigue of material properties and crack merger, leading to a transition from slow to rapid crack growth rate. Use of these models can identify critical C-rates and particle sizes which mitigate cracking.At the continuum scale, we then investigate how these stresses evolve during cycling, with stress heterogeneities arising initially at the electrode-separator interface, but later propagating to the electrode-current collector interface [2]. We then extend this study to composite anodes of graphite and silicon, where highly non-linear behaviour is observed [3]. Here, the graphite phase provides the majority of the reaction current density at high state-of-charge operation, with this then shifting to the silicon phase at low state-of-charge operation, with hysteresis effects observed due to the silicon. This behaviour is attributed mostly to the different open circuit potentials of the two different phases. Finally, we explore how these effects propagate to the battery pack scale and highlight the divergence of material level performance from their real-world implementation [4].

Multi-scale feature fusion model for real-time Blood glucose monitoring and hyperglycemia prediction based on wearable devices.

Ship hydrodynamic performance evaluation technique is one of the prerequisites and key conditions for ship-type optimization based on SBD technology, which provides an effective means to establish the mathematical model of ship design optimization problem. The accuracy of the hydrodynamic performance evaluation directly affects the quality of the optimization results [1]. In the optimization process, the optimization algorithm will adjust the next search direction based on the performance prediction results. Therefore, the reliability of the performance prediction results is the key to ensure that the search direction of the optimization algorithm in the design space is correct or not, and is also directly related to the success or failure of the optimization design.

In this paper, the problem of parallel loop scheduling for heterogeneous networks of workstations is discussed. A practical network of workstations is heterogeneous where computing power varies in different composing workstations. A simple loop scheduling technique is insufficient in exploiting maximum computing power of such a system. We propose a fundamental idea for a performance prediction tool to gauge the relative computing power among composing workstations so that parallel performance of a program run on a given network can be predicted. In view of the performance prediction results, a loop scheduling approach is then incorporated into the system to achieve close-to-optimal parallel performance. Examples running benchmark programs demonstrate a significant gain from the proposed approach over traditional scheduling approaches.

The Fast-Steering secondary mirror (FSM) of Giant Magellan Telescope (GMT) consists of seven 1.1m diameter segments with effective diameter of 3.2m. Each of the seven segments is held by three axial supports and a central lateral support with a vacuum system to mirror gravity compensations. The mirror assembly has a tip-tilt feature capable of guiding to attenuate telescope wind shake and mount control jitter. Both on-axis and off-axis mirror segments were optimized under various design considerations using finite element analyses and optical performance analyses. An integrated model of the mirror cell assembly was developed including sub-assembly parts consisting of axial supports, lateral support, breakaway mechanism, seismic restraints, and pressure seal. In this paper for the optimized FSM mirror and mirror cell assembly, the design considerations are addressed, and performance prediction results are discussed in detail with respect to the specifications.

A computation scheme was implemented in a general-purpose, finite-difference simulator to model multi-lateral wells. Frictional pressure drop along the wellbore and proper fluid mixing at lateral connection points were included in the calculation of the wellbore pressure profile for each lateral. The formulation is general and a lateral can be branched out from any point along the main well path. Since the production profile of each lateral is influenced by the wellbore pressure distribution, the prediction accuracy depends on a good wellbore flow model. The Beggs and Brill(1) correlation and the homogeneous flow model were implemented, along with a new model proposed by Ouyang et al.(2) which included an acceleration term and accounted for the lubrication effect due to radial influx. Well performance prediction results using the three models were compared. The impacts of different tubing sizes on the well performance and the production contribution from each lateral were studied. Introduction Recently multi-lateral wells have been employed in off-shore platforms or remote and hard to access terrain areas to save costs comparing with drilling multiple vertical or horizontal wells. This work is to study the pressure and production profiles along each lateral using different pipe flow models. The vertical standoffs of laterals to GOC or WOC, the optimal length of laterals, the angles between laterals, the completion PI values used in simulation, and the optimal number of laterals are other important factors in designing multi-lateral wells. However, these issues will not be addressed here. Numerical Method The completion production rate and wellbore pressure profiles along each lateral are closely related. There is also interaction between laterals. The fluid dynamics calculation inside the wellbore can be described as follows: Pipeflow Calculation There are many steady-state correlations available to calculate frictional pressure losses. In this work, we implemented three models, namely, the Beggs and Brill model, the homogeneous model, and a modified homogeneous model (including radial inflow effect) in a general-purpose reservoir simulator. The homogeneous model gave a better pressure match for the Stanford/Marathon two-phase flow tests(3,4). It was a very simplified model. For two-phase flow, it used a mixing rule to calculate the density and viscosity values of the gas-liquid mixture. Then it used the mixture properties in the single-phase fluid-wall frictional loss formula. We did not use a transient pipeflow model(5) because it would require much longer computing time. Radial Inflow Effect Rigorously speaking, the regular pipe flow calculation is not suitable for open hole or perforation intervals due to the presence of radial inflow from the reservoir. For turbulent flow, the radial inflow provides a so called "lubrication effect" and reduces the frictional loss(6,7). For laminar flow (rarely occurring in oil field practice), it increases the frictional loss(8). Here we used a formula proposed by Ouyang et al.(3) to model radial inflow effect, Equation (1) (Available In Full Paper) where ƒo is the corrected frictional coefficient due to radial inflow and ƒ is the wall friction coefficient. For the open hole, Rew can be expressed as Equation (2) (Available In Full Paper)

Research problems in the domains of physical, engineering, biological sciences often span multiple time and length scales, owing to the complexity of information transfer underlying mechanisms. Multiscale modeling (MSM) and high‐performance computing (HPC) have emerged as indispensable tools for tackling such complex problems. We review the foundations, historical developments, and current paradigms in MSM. A paradigm shift in MSM implementations is being fueled by the rapid advances and emerging paradigms in HPC at the dawn of exascale computing. Moreover, amidst the explosion of data science, engineering, and medicine, machine learning (ML) integrated with MSM is poised to enhance the capabilities of standard MSM approaches significantly, particularly in the face of increasing problem complexity. The potential to blend MSM, HPC, and ML presents opportunities for unbound innovation and promises to represent the future of MSM and explainable ML that will likely define the fields in the 21st century.

There is mounting experimental evidence that network traffic processes exhibit ubiquitous properties of self-similarity and long range dependence (LRD), i.e. of correlations over a wide range of time scales. However, there is still considerable debate about how to model such processes and about their impact on network and application performance. In this paper, we argue that much recent modeling work has failed to consider the impact of two important parameters, namely the finite range of time scales of interest in performance evaluation and prediction problems, and the first-order statistics such as the marginal distribution of the process.We introduce and evaluate a model in which these parameters can be easily controlled. Specifically, our model is a modulated fluid traffic model in which the correlation function of the fluid rate is asymptotically second-order self-similar with given Hurst parameter, then drops to zero at a cutoff time lag. We develop a very efficient numerical procedure to evaluate the performance of the single server queue fed with the above fluid input process. We use this procedure to examine the fluid loss rate for a wide range of marginal distributions, Hurst parameters, cutoff lags, and buffer sizes.Our main results are as follows. First, we find that the amount of correlation that needs to be taken into account for performance evaluation depends not only on the correlation structure of the source traffic, but also on time scales specific to the system under study. For example, the time scale associated to a queueing system is a function of the maximum buffer size. Thus for finite buffer queues, we find that the impact on loss of the correlation in the arrival process becomes nil beyond a time scale we refer to as the correlation horizon. Second, we find that loss depends in a crucial way on the marginal distribution of the fluid rate process. Third, our results suggest that reducing loss by buffering is hard. We advocate the use of source traffic control and statistical multiplexing instead.

This study analysed Culicoides presence-absence data from 46 sampling sites in Germany, where monitoring was carried out from April 2007 until May 2008. Culicoides presence-absence data were analysed in relation to land cover data, in order to study whether the prevalence of biting midges is correlated to land cover data with respect to the trapping sites. We differentiated eight scales, i.e. buffer zones with radii of 0.5, 1, 2, 3, 4, 5, 7.5 and 10 km, around each site, and chose several land cover variables. For each species, we built eight single-scale models (i.e. predictor variables from one of the eight scales for each model) based on averaged, generalised linear models and two multiscale models (i.e. predictor variables from all of the eight scales) based on averaged, generalised linear models and generalised linear models with random forest variable selection. There were no significant differences between performance indicators of models built with land cover data from different buffer zones around the trapping sites. However, the overall performance of multi-scale models was higher than the alternatives. Furthermore, these models mostly achieved the best performance for the different species using the index area under the receiver operating characteristic curve. However, as also presented in this study, the relevance of the different variables could significantly differ between various scales, including the number of species affected and the positive or negative direction. This is an even more severe problem if multi-scale models are concerned, in which one model can have the same variable at different scales but with different directions, i.e. negative and positive direction of the same variable at different scales. However, multi-scale modelling is a promising approach to model the distribution of Culicoides species, accounting much more for the ecology of biting midges, which uses different resources (breeding sites, hosts, etc.) at different scales.

This paper presents a new and simple search technique to determine the optimum control parameters, such as turn-on and turn-off angles, current set level, etc., of a switched reluctance drive. Thanks to this new technique, several drive performance quantities such as efficiencies, torque ripple, energy consumption, torque per ampere, etc., can be optimized simultaneously over the entire torque- or power-speed operation range of the drive. In prior work, these performance quantities were sequentially optimized, one at a time, this despite reported evidence that improving one performance quantity, such as efficiency, can have a negative impact on others, such as torque ripple. The technique was successfully used offline to determine the optimum firing angles for achieving highest drive efficiency with lower torque ripple in a 102-kW switched reluctance motor drive intended for electric vehicle propulsion. The performance prediction and test results are found in good correlation.