Search NASA⌕ Search

SEARCH · Search NASA

Results for “Dynamic load balancing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 379 records · Page 21

Efficient Machine Learning Approach for Optimizing Scientific Computing Applications on Emerging HPC Architectures

Efficient parallel implementations of scientific applications on multi-core CPUs with accelerators such as GPUs and Xeon Phis is challenging. This requires - exploiting the data parallel architecture of the accelerator along with the vector pipelines of modern x86 CPU architectures, load balancing, and efficient memory transfer between different devices. It is relatively easy to meet these requirements for highly structured scientific applications. In contrast, a number of scientific and engineering applications are unstructured. Getting performance on accelerators for these applications is extremely challenging because many of these applications employ irregular algorithms which exhibit data-dependent control-ow and irregular memory accesses. Furthermore, these applications are often iterative with dependency between steps, and thus making it hard to parallelize across steps. As a result, parallelism in these applications is often limited to a single step. Numerical simulation of charged particles beam dynamics is one such application where the distribution of work and memory access pattern at each time step is irregular. Applications with these properties tend to present significant branch and memory divergence, load imbalance between different processor cores, and poor compute and memory utilization. Prior research on parallelizing such irregular applications have been focused around optimizing the irregular, data-dependent memory accesses and control-ow during a single step of the application independent of the other steps, with the assumption that these patterns are completely unpredictable. We observed that the structure of computation leading to control-ow divergence and irregular memory accesses in one step is similar to that in the next step. It is possible to predict this structure in the current step by observing the computation structure of previous steps. In this dissertation, we present novel machine learning based optimization techniques to address the parallel implementation challenges of such irregular applications on different HPC architectures. In particular, we use supervised learning to predict the computation structure and use it to address the control-ow and memory access irregularities in the parallel implementation of such applications on GPUs, Xeon Phis, and heterogeneous architectures composed of multi-core CPUs with GPUs or Xeon Phis. We use numerical simulation of charged particles beam dynamics simulation as a motivating example throughout the dissertation to present our new approach, though they should be equally applicable to a wide range of irregular applications. The machine learning approach presented here use predictive analytics and forecasting techniques to adaptively model and track the irregular memory access pattern at each time step of the simulation to anticipate the future memory access pattern. Access pattern forecasts can then be used to formulate optimization decisions during application execution which improves the performance of the application at a future time step based on the observations from earlier time steps. In heterogeneous architectures, forecasts can also be used to improve the memory performance and resource utilization of all the processing units to deliver a good aggregate performance. We used these optimization techniques and anticipation strategy to design a cache-aware, memory efficient parallel algorithm to address the irregularities in the parallel implementation of charged particles beam dynamics simulation on different HPC architectures. Experimental result using a diverse mix of HPC architectures shows that our approach in using anticipation strategy is effective in maximizing data reuse, ensuring workload balance, minimizing branch and memory divergence, and in improving resource utilization.

Arumugam, Kamesh↗

Recovery Act: Federspiel Controls (now Vigilent) and State of California Department of General Services Data Center Energy Efficient Cooling Control Demonstration. Final technical project report

Eight State of California data centers were equipped with an intelligent energy management system to evaluate the effectiveness, energy savings, dollar savings and benefits that arise when powerful artificial intelligence-based technology measures, monitors and actively controls cooling operations. Control software, wireless sensors and mesh networks were used at all sites. Most sites used variable frequency drives as well. The system dynamically adjusts temperature and airflow on the fly by analyzing real-time demands, thermal behavior and historical data collected on site. Taking into account the chaotic interrelationships of hundreds to thousands of variables in a data center, the system optimizes the temperature distribution across a facility while also intelligently balancing loads, outputs, and airflow. The overall project will provide a reduction in energy consumption of more than 2.3 million kWh each year, which translates to $240,000 saved and a reduction of 1.58 million pounds of carbon emissions. Across all sites, the cooling energy consumption was reduced by 41%. The average reduction in energy savings across all the sites that use VFDs is higher at 58%. Before this case study, all eight data centers ran the cooling fans at 100% capacity all of the time. Because of the new technology, cooling fans run at the optimum fan speed maintaining stable air equilibrium while also expending the least amount of electricity. With lower fan speeds, the life of the capital investment made on cooling equipment improves, and the cooling capacity of the data center increases. This case study depicts a rare technological feat: The same process and technology worked cost effectively in eight very different environments. The results show that savings were achieved in centers with diverse specifications for the sizes, ages and types of cooling equipment. The percentage of cooling energy reduction ranged from 19% to 78% while keeping temperatures substantially within the limits recommended by the American Society of Heating, Refrigeration, and Air-Conditioning Engineers (ASHRAE) for data center facilities.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Evaluating In-Clique and Topological Parallelism Strategies for Junction Tree-Based Bayesian Inference Algorithm on the Cray XMT

Long viewed as a strong statistical inference technique, Bayesian networks have emerged to be an important class of applications for high-performance computing. We have applied an architecture-conscious approach to parallelizing the Lauritzen-Spiegelhalter Junction Tree algorithm for exact inferencing in Bayesian networks. In optimizing the Junction Tree algorithm, we have implemented both in-clique and topological parallelism strategies to best leverage the fine-grained synchronization and massive-scale multithreading of the Cray XMT architecture. Two topological techniques were developed to parallelize the evidence propagation process through the Bayesian network. One technique involves performing intelligent scheduling of junction tree nodes based on its topology and relative size. The second technique involves decomposing the junction tree into a much finer tree-like representation to offer much more opportunities for parallelism. We evaluate these optimizations on five different Bayesian networks and report our findings and observations. Another important contribution of this paper is to demonstrate the application of massive-scale multithreading for load balancing and use of implicit parallelism-based compiler optimizations in designing scalable inferencing algorithms.

99 GENERAL AND MISCELLANEOUS↗

The influence of dynamic inflow and torsional flexibility on rotor damping in forward flight from symbolically generated equations

The combined effects of blade torsion and dynamic inflow on the aeroelastic stability of an elastic rotor blade in forward flight are studied. The Helicopter Equations for Stability and Loads (HESL) program is extended to derive the governing equations of motion for the blade, and a Lagrangian formulation is used to obtain the equations in generalized coordinates. The program generates the steady-state and linearized perturbation equations in symbolic form and then codes them into FORTRAN subroutines. The coefficients for each equation and for each mode are identified through a numerical program; the latter can also be used to obtain the harmonic balance equations. The governing multiblade equations are derived explicitly using HESL. These equations can accommodate any number of elastic blade modes. Stability results are presented for several hingeless rotor blade structural models, and the influence of dynamic inflow in forward flight with an elastic hingeless rotor is investigated.

Reddy, T. S. R.↗

Rendezvous algorithms for large-scale modeling and simulation

Rendezvous algorithms encode a communication pattern that is useful when processors sending data do not know who the receiving processors should be, or vice versa. The idea is to define an intermediate decomposition where datums from different sending processors can ”rendezvous” to perform a computation, in a manner that both the senders and eventual receivers of the results can identify the appropriate rendezvous processor. Though they were originally designed for interpolating between overlaid grids with independent parallel decompositions (Plimpton et al., 2004), we have recently found rendezvous algorithms useful for a variety of operations in particle- or grid-based simulation codes when running large problems on large numbers of processors. In particular, we show they can perform well when a load-balanced intermediate decomposition is randomized and not spatial, requiring all-to-all communication to move data between processors. In this case rendezvous algorithms leverage the large bisection communication bandwidths which parallel machines provide. We describe how rendezvous algorithms work in a scientific computing context and give specific examples for molecular dynamics and Direct Simulation Monte Carlo codes which result in dramatic performance improvements versus simpler algorithms which do not scale as well. We explain how a generic rendezvous algorithm can be implemented, and also point out similarities with the MapReduce paradigm popularized by Google and Hadoop.

97 MATHEMATICS AND COMPUTING↗

A Latency-Tolerant Partitioner for Distributed Computing on the Information Power Grid

NASA's Information Power Grid (IPG) is an infrastructure designed to harness the power of graphically distributed computers, databases, and human expertise, in order to solve large-scale realistic computational problems. This type of a meta-computing environment is necessary to present a unified virtual machine to application developers that hides the intricacies of a highly heterogeneous environment and yet maintains adequate security. In this paper, we present a novel partitioning scheme. called MinEX, that dynamically balances processor workloads while minimizing data movement and runtime communication, for applications that are executed in a parallel distributed fashion on the IPG. We also analyze the conditions that are required for the IPG to be an effective tool for such distributed computations. Our results show that MinEX is a viable load balancer provided the nodes of the IPG are connected by a high-speed asynchronous interconnection network.

Das, Sajal K.↗

Floating wind turbine system

A floating wind turbine system with a tower structure that includes at least one stability arm extending therefrom and that is anchored to the sea floor with a rotatable position retention device that facilitates deep water installations. Variable buoyancy for the wind turbine system is provided by buoyancy chambers that are integral to the tower itself as well as the stability arm. Pumps are included for adjusting the buoyancy as an aid in system transport, installation, repair and removal. The wind turbine rotor is located downwind of the tower structure to allow the wind turbine to follow the wind direction without an active yaw drive system. The support tower and stability arm structure is designed to balance tension in the tether with buoyancy, gravity and wind forces in such a way that the top of the support tower leans downwind, providing a large clearance between the support tower and the rotor blade tips. This large clearance facilitates the use of articulated rotor hubs to reduced damaging structural dynamic loads. Major components of the turbine can be assembled at the shore and transported to an offshore installation site.

Viterna, Larry A.↗

Simulation of dynamic fracture using peridynamics, finite element modeling, and contact.

Peridynamics is a nonlocal extension of classical solid mechanics that allows for the modeling of bodies in which discontinuities occur spontaneously. Because the peridynamic expression for the balance of linear momentum does not contain spatial derivatives and is instead based on an integral equation, it is well suited for modeling phenomena involving spatial discontinuities such as crack formation and fracture. In this study, both peridynamics and classical finite element analysis are applied to simulate material response under dynamic blast loading conditions. A combined approach is utilized in which the portion of the simulation modeled with peridynamics interacts with the finite element portion of the model via a contact algorithm. The peridynamic portion of the analysis utilizes an elastic-plastic constitutive model with linear hardening. The peridynamic interface to the constitutive model is based on the calculation of an approximate deformation gradient, requiring the suppression of possible zero-energy modes. The classical finite element portion of the model utilizes a Johnson-Cook constitutive model. Simulation results are validated by direct comparison to expanding tube experiments. The coupled modeling approach successfully captures material response at the surface of the tube and the emerging fracture pattern. The coupling of peridynamics and finite element analysis via a contact algorithm has been shown to be a viable means for simulating material fracture in a high-velocity impact experiment. A combined peridynamics/finite element approach was applied to model an expanding tube experiment performed by Vogler, et al., in which loading on the tube is a result of Lexan slugs impacting inside the tube. The Lexan portion of the simulation was modeled with finite elements and a Johnson-Cook elastic-plastic material model in conjunction with an equation-of-state law. The steel tube portion of the simulation was modeled with peridynamics, an elastic-plastic material model, and a critical stretch bond damage model. The application of peridynamics to the tube portion of the model allowed the capture of the formation of cracks and eventual fragmentation of the tube. The simulation results yielded good agreement with the experimental results published by Vogler, et al., for the velocity and displacement profiles on the surface of the tube and the resulting fragment distribution. Numerical difficulties were encountered that required removal of hexahedron elements from the Lexan portion of the model over the course of the simulation. The significant number of inverted and nearly-inverted elements appearing over the course of the simulation is believed to be a result of irregularities in the contact between the Lexan and AerMet portions of the model, and was likely exacerbated by the ultra-high strength of the AerMet tube. Future simulations are planned in which the Lexan portion of the simulation is modeled with peridynamics, or with an alternative method such as smoothed particle hydrodynamics, with the goal of reducing these numerical difficulties.

36 MATERIALS SCIENCE↗

Analysis of ISO NE Balancing Requirements: Uncertainty-based Secure Ranges for ISO New England Dynamic Inerchange Adjustments

The document describes detailed uncertainty quantification (UQ) methodology developed by PNNL to estimate secure ranges of potential dynamic intra-hour interchange adjustments in the ISO-NE system and provides description of the dynamic interchange adjustment (DINA) tool developed under the same contract. The overall system ramping up and down capability, spinning reserve requirements, interchange schedules, load variations and uncertainties from various sources that are relevant to the ISO-NE system are incorporated into the methodology and the tool. The DINA tool has been tested by PNNL and ISO-NE staff engineers using ISO-NE data.

Etingov, Pavel V.↗

Aeroservoelastic Control Law Development for the Integrated Adaptive Wing Technology Maturation Wind-Tunnel Test

The NASA Advanced Air Transport Technology (AATT) Project seeks to “explore and develop technologies and concepts for improved energy efficiency and environmental compatibility for fixed wing subsonic transports” [1]. The IAWTM project is a subproject within AATT and is a joint Boeing/NASA effort to, in part, demonstrate the active control systems necessary to employ high-aspect ratio wings for improved aerodynamic efficiency. The aircraft configuration under study for the IAWTM project is a generic jet transport based on the NASA Common Research Model (CRM) whose wing aspect ratio has been increased to 13.5 (as opposed to nominal CRM aspect ratio of 9) and whose development is given in Ref. [2]. A semi-span model of the high-aspect-ratio CRM with 10 trailing edge control surfaces is currently being fabricated and will be delivered to the NASA Langley Transonic Dynamics Tunnel (TDT). There, several tests will be conducted to demonstrate maneuver load alleviation (MLA), gust load alleviation (GLA), active flutter suppression (AFS), and drag reduction, all at transonic conditions. A layout of the wind tunnel model is shown in Fig. 1. The wind-tunnel model is a spar/pod wing design with a relatively rigid fuselage attached to a mounting structure (seen in Fig. 1), which in turn is attached to a load balance behind the tunnel wall. Shims exist that can be attached to force the load to go through the fuselage rather than through the balance so as not to damage the balance in the event of flutter. The spar/pod design is similar to that used in the Boeing truss-braced wing aeroelastic model tested in the TDT in 2013 [3]. There is a central spar running the span of the wing to which the pods are attached that form the aerodynamic shape of the wing. The pods are constructed of ribs for attachment to the spar and skins to create the outer mold line. There are ten active control surfaces that will be used for the test objectives discussed above. Three of these (two outboard ailerons and one inboard aileron) are high-speed electrohydraulic actuators, and the remaining seven are low-speed mini-plane flaps, which have slower servo motors that will be used primarily for drag and maneuver load reductions. Control laws must be developed for each test (MLA, GLA, AFS, and drag reduction) and must be able to work in tandem when two or more control strategies are being tested simultaneously (e.g., when control laws to mitigate maneuver loads and gust loads operate simultaneously). The development of these control laws requires accurate mathematical representations of the aerodynamics, structural dynamics, and control surface dynamics of the system. Taken together, this mathematical representation—which is often in state-space form— is the ASE model. Three low-order ASE models of differing fidelities were developed for the IAWTM test and will be presented here. The differences in fidelity appeared in the aerodynamic modeling—one model assumed purely linear aerodynamics, while the other two models used different techniques to linearize the aerodynamics about a nonlinear mean. Furthermore, simple observer-based full-state feedback controllers for each test (MLA, GLA, AFS, and drag reduction) were designed from these low-order ASE models and implemented in a higher-order, nonlinear computational ASE simulation to assess the effectiveness of each controller. Observers were designed from each of the three lower-order ASE models, and each (linear) observer was then used as an estimate of the nonlinear dynamics of the system. Using the estimated states and the full-state feedback controller, control surface commands were fed back to the higher-order, nonlinear ASE system. The insight gained during this development will be applied to control law development for the physical IAWTM wind-tunnel test.

Josiah M Waite↗

Studying effects of non-equilibrium radiative transfer via HPC

This report presents slides on Ph.D. Research Goals; Local Thermodynamic Equilibrium (LTE) Implications; Calculating an Opacity; Opacity: Pictographic Representation; Opacity: Pictographic Representation; Opacity: Pictographic Representation; Collisional Radiative Modeling; Radiative and Collisional Excitation; Photo and Electron Impact Ionization; Autoionization; The Rate Matrix; Example: Total Photoionization rate; The Rate Coefficients; inlinlte version 1.1; inlinlte: Verification; New capabilities: Rate Matrix – Flexibility; Memory Option Comparison; Improvements over previous DCA solver; Inter- and intra-node load balancing; Load Balance – Full Picture; Load Balance – Full Picture; Load Balance – Internode; Load Balance – Scaling; Description; Performance; xRAGE Simulation; Post-process @ 2hr; Post-process @ 4hr; Post-process @ 8hr; Takeaways; Performance for 1 realization; Motivation for QOI; Multigroup Er; Transport and NLTE large effects (1mm, 1keV); Transport large effect, NLTE lesser (1mm, 750eV); Blastwave Diagnostici – Description & Performance; Temperature Comparison; NLTE has effect on dynamics at wall; NLTE has lesser effect in the foam; Global Takeaways; The end.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Pick-up ion pressure gradients modulating the solar wind dynamics

Neutral interstellar atoms penetrate deeply into the inner heliosphere before they become ionized by various processes. As ions they are picked-up by the frozen-in magnetic fields and are convected outwards with the solar wind plasma. Thereby the primary plasma flow is mass, momentum, and energy-loaded. The dynamics of the distant multi-constituent solar wind is, however, not solely determined by these loading processes, but is also affected by the wave-mediated pick-up ion pressure gradients derivable from the pick-up ion distribution function. The action of the radial components of these pressures essentially counter balances the decelerating effect of the solar wind momentum loading, diminishing strongly the deceleration of the distant solar wind. Furthermore the latitudinal components of the pick-up ion pressures induce latitudinal forces acting on the multiconstituent solar plasma outflow and inducing nonradial bulk flow components. The enforced nonradial outflow geometry on the upwind hemisphere may partly be responsible for the magnetic flux deficit which was claimed since several years in the PIONEER-10 magnetic flux data.

Fahr, Hans J.↗

Tri-Level Linear Programming Model for Automatic Load Shedding Using Spectral Clustering

Traditional load shedding schemes can be inadequate in grids with high renewable penetration, leading to unstable events and unnecessary grid islanding. Although for both manual and automatic operating modes load shedding areas have been predefined by grid operators, they have remained fixed, and may be sub-optimal due to dynamic operating conditions. In this work, a distributed tri-level linear programming model for automatic load shedding to avoid system islanding is presented. Preventing islanding is preferred because it reduces the need for additional load shedding besides the disconnection of transmission lines between islands. This is crucial as maintaining the local generation-demand balance is necessary to preserve frequency stability. Furthermore, uneven distribution of generation resources among islands can lead to increased load shedding, causing economic and reliability challenges. This issue is further compounded in modern power systems heavily dependent on non-dispatchable resources like wind and solar. The upper-level model uses complex power flow measurements to determine the system areas to shed load depending on actual operating conditions using a spectral clustering approach. The mid-level model estimates the area system state, while the lower-level model determines the locations and load values to be shed. The solution is practical and promising for real-world applications.

Baquedano-Aguilar, Mario D.↗

Variable Acceleration Force Calibration System (VACS)

Conventionally, force balances have been calibrated manually, using a complex system of free hanging precision weights, bell cranks, and/or other mechanical components. Conventional methods may provide sufficient accuracy in some instances, but are often quite complex and labor-intensive, requiring three to four man-weeks to complete each full calibration. To ensure accuracy, gravity-based loading is typically utilized. However, this often causes difficulty when applying loads in three simultaneous, orthogonal axes. A complex system of levers, cranks, and cables must be used, introducing increased sources of systematic error, and significantly increasing the time and labor intensity required to complete the calibration. One aspect of the VACS is a method wherein the mass utilized for calibration is held constant, and the acceleration is changed to thereby generate relatively large forces with relatively small test masses. Multiple forces can be applied to a force balance without changing the test mass, and dynamic forces can be applied by rotation or oscillating acceleration. If rotational motion is utilized, a mass is rigidly attached to a force balance, and the mass is exposed to a rotational field. A large force can be applied by utilizing a large rotational velocity. A centrifuge or rotating table can be used to create the rotational field, and fixtures can be utilized to position the force balance. The acceleration may also be linear. For example, a table that moves linearly and accelerates in a sinusoidal manner may also be utilized. The test mass does not have to move in a path that is parallel to the ground, and no re-leveling is therefore required. Balance deflection corrections may be applied passively by monitoring the orientation of the force balance with a three-axis accelerometer package. Deflections are measured during each test run, and adjustments with respect to the true applied load can be made during the post-processing stage. This paper will present the development and testing of the VASC concept.

Rhew, Ray D.↗

Damping Effects of Drogue Parachutes on Orion Crew Module Dynamics

Because simulations of the Orion Crew Module (CM) dynamics with drogue parachutes deployed were under-predicting the amount of damping seen in free-flight tests, an attach-point damping model was applied to the Orion system. A key hypothesis in this model is that the drogue parachutes' net load vector aligns with the CM drogue attachment point velocity vector. This assumption seems reasonable and has historically produced good results, but has never been experimentally verified. The wake of the CM influences the drogue parachutes, which makes performance predictions of the parachutes difficult. Many of these effects are not currently modeled in the simulations. A forced oscillation test of the CM with parachutes was conducted in the NASA LaRC 20-Ft Vertical Spin Tunnel (VST) to gather additional data to validate and refine the attach-point damping model. A second loads balance was added to the original Orion VST model to measure the drogue parachute loads independently of the CM. The objective of the test was to identify the contribution of the drogues to CM damping and provide additional information to quantify wake effects and the interactions between the CM and parachutes. The drogue parachute force vector was shown to be highly dependent on the CM wake characteristics. Based on these wind tunnel test data, the attach-point damping model was determined to be a sufficient approximation of the parachute dynamics in relationship to the CM dynamics for preliminary entry vehicle system design. More wake effects should be included to better model the system.

Aubuchon, Vanessa V.↗

Utilization of Model Predictive Control to Balance Power Absorption Against Load Accumulation

Wave energy converter (WEC) control strategies have been primarily focused on maximizing power absorption. The use of model predictive control strategies allows for a finite-horizon, multiterm objective function to be solved. This work utilizes a multiterm objective function to maximize power absorption while minimizing the structural loads on the WEC system. Furthermore, a Kalman filter and autoregressive model were used to estimate and forecast the wave exciting force and predict the future dynamics of the WEC. The WEC's power-take-off time-averaged power and structural loads under a perfect forecast assumption in irregular waves were compared against results obtained from the Kalman filter and autoregressive model to evaluate model predictive control performance.

16 TIDAL AND WAVE POWER↗

Modeling of Electric Water Heaters for Demand Response: A Baseline PDE Model

Demand response (DR)control can effectively relieve balancing and frequency regulation burdens on conventional generators, facilitate integrating more renewable energy, and reduce generation and transmission investments needed to meet peak demands. Electric water heaters (EWHs) have a great potential in implementing DR control strategies because: (a) the EWH power consumption has a high correlation with daily load patterns; (b) they constitute a significant percentage of domestic electrical load; (c) the heating element is a resistor, without reactive power consumption; and (d) they can be used as energy storage devices when needed. Accurately modeling the dynamic behavior of EWHs is essential for designing DR controls. Various water heater models, simplified to different extents, were published in the literature; however, few of them were validated against field measurements, which may result in inaccuracy when implementing DR controls. In this paper, a partial differential equation physics-based model, developed to capture detailed temperature profiles at different tank locations, is validated against field test data for more than 10 days. The developed model shows very good performance in capturing water thermal dynamics for benchmark testing purposes

Balancing service, demand response, electric water↗

Utilization of Model Predictive Control to Balance Power Absorption Against Load Accumulation: Preprint

Wave energy converter (WEC) control strategies have been primarily focused on maximizing power absorption. The use of model predictive control strategies allows for a finite-horizon, multiterm objective function to be solved. This work utilizes a multiterm objective function to maximize power absorption while minimizing the structural loads on the WEC system. Furthermore, a Kalman filter and autoregressive model were used to estimate and forecast the wave exciting force and predict the future dynamics of the WEC. The WEC's power-take-off time-averaged power and structural loads under a perfect forecast assumption in irregular waves were compared against results obtained from the Kalman filter and autoregressive model to evaluate model predictive control performance.

16 TIDAL AND WAVE POWER↗