Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “performance tuning programming systems”

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 37 records · Page 2

Long-Term Carbon Injection Field Test for 90% Mercury Removal for a PRB Unit a Spray Dryer and Fabric Filter

The power industry in the U.S. is faced with meeting regulations to reduce the emissions of mercury compounds from coal-fired plants. Injecting a sorbent such as powdered activated carbon (PAC) into the flue gas represents one of the simplest and most mature approaches to controlling mercury emissions from coal-fired boilers. The purpose of this test program was to evaluate the long-term mercury removal capability, long-term mercury emissions variability, and operating and maintenance (O&M) costs associated with sorbent injection on a configuration being considered for many new plants. Testing was conducted by ADA Environmental Solutions (ADA) at Rocky Mountain Power’s (RMP) Hardin Station through funding provided by DOE/NETL, RMP, and other industry partners. The Hardin Station is a new plant rated at 121 MW gross that was first brought online in April of 2006. Hardin fires a Powder River Basin (PRB) coal and is configured with selective catalytic reduction (SCR) for NOx control, a spray dryer absorber (SDA) for SO2 control, and a fabric filter (FF) for particulate control. Based upon previous testing at PRB sites with SCRs, very little additional mercury oxidation from the SCR was expected at Hardin. In addition, based upon results from DOE/NETL Phase II Round I testing at Holcomb Station and results from similarly configured sites, low native mercury removal was expected across the SDA and FF. The main goal of this project was met—sorbent injection was used to economically and effectively achieve 90% mercury control as measured from the air heater (AH) outlet to the stack for a period of ten months. This goal was achieved with DARCO® Hg-LH, Calgon FLUEPAC®-MC PLUS and ADA Power PAC PREMIUM brominated activated carbons at nominal loadings of 1.5–2.5 lb/MMacf. An economic analysis determined the twenty-year levelized cost to be 0.87 mills/kW-hr, or $15,000/lb Hg removed. No detrimental effects on other equipment or plant operations were observed. The results of this project also filled a data gap for plants firing PRB coal and configured with an SCR, SDA, and FF, as many new plants are being designed today. Another goal of the project was to evaluate, on a short-term basis, the mercury removal associated with coal additives and coal blending with western bituminous coal. The additive test showed that, at this site, the coal additive known as KNX was affective at increasing mercury removal while decreasing sorbent usage. Coal blending was conducted with two different western bituminous coals, and West Elk coal increased native capture from nominally 10% to 50%. Two additional co-benefits were discovered at this site. First, it was found that native capture increased from nominally 10% at full load to 50% at low load. The effect is believed to be due to an increase in mercury oxidation across the SCR caused by a corresponding decrease in ammonia injection when the plant reduces load. Less ammonia means more active oxidation sites in the SCR for the mercury. The second co-benefit was the finding that high ammonia concentrations can have a negative impact on mercury removal by powdered activated carbon. For a period of time, the plant operated with a high excess of ammonia injection necessitated by the plugging of one-third of the SCR. Under these conditions and at high load, the mercury control system could not maintain 90% removal even at the maximum feed rate of 3.5 lb/MMacf (pounds of mercury per million actual cubic feet). The plant was able to demonstrate that mercury removal was directly related to the ammonia injection rate in a series of tests where the ammonia rate was decreased, causing a corresponding increase in mercury removal. Also, after the SCR was refurbished and ammonia injection levels returned to normal, the mercury removal performance also returned to normal. Another goal of the project was to install a commercial-grade activated carbon injection (ACI) system and integrate it with new-generation continuous emissions monitors for mercury (Hg-CEMs) to allow automatic feedback control on outlet mercury emissions. This was accomplished and the plant can now be operated to control carbon injection based on either the overall mercury removal or an outlet mercury emission rate. By integrating these systems, it was determined that the plant could reduce powdered activated carbon consumption, especially at low load, because, at Hardin, native mercury capture increases from less than 20% to about 50% at low load and the carbon injection rate can be decreased accordingly. Currently, the plant is operating to automatically control emissions to below 0.9 lb/TBtu (pounds of mercury per million British thermal units) at carbon loadings of 0.5 to 1.5 lb/MMacf. During the final phase of the Long-Term test, the ACI system was operated by plant personnel. The estimated O&M cost for a single Hg-CEM system is $15,500/yr. The Hg-CEMs performed well throughout the project. This project began shortly after Thermo Fisher first offered the Mercury Freedom System™ on a commercial basis and progressed though several iterations, improvements, and upgrades to the hardware and software. Indeed, there was a ten-fold increase in the precision and accuracy of the units during the course of the project due to several successful upgrades. In their present condition, the Hg-CEMs measure mercury to a precision of about ± 0.05 μg/wscm (micrograms of mercury per wet standard cubic meter of gas), and only require occasional fine-tuning of the calibration coefficients. The quality assurance/quality control (QA/QC) protocol required to keep the units operating at their optimal performance was also developed and perfected during the course of the project. ADA Environmental Solutions (ADA) developed a daily calibration procedure that surpasses the requirements specified in the Clean Air Mercury Rule (CAMR), and a weekly diagnostic program that ensures that the systems are operating properly and receive the necessary maintenance. For the most part, the systems passed the daily, weekly, and quarterly QA/QC requirements as well as four performance verification tests using the Ontario Hydro (O-H) and Sorbent Trap Methods (STM) for the first test and the EPA Method 30A (M30A) procedure for the remaining three. However, some improvements are still necessary before the system can meet all of the requirements. These involve tests that challenge the system with oxidized mercury (Hg+2). These tests could not be passed at Hardin in spite of trying several improvements suggested by ADA or Thermo Fisher.

20 FOSSIL-FUELED POWER PLANTS↗

Managing code transformations for better performance portability

Code optimization is an intricate task that is getting more complex as computing systems evolve. Managing the program optimization process, including the implementation and evaluation of code variants, is tedious, inefficient, and errors are likely to be introduced in the process. Moreover, because each platform typically requires a different sequence of transformations to fully harness its computing power, the optimization process complexity grows as new platforms are adopted. To address these issues, systems and frameworks have been proposed to automate the code optimization process. They, however, have not been widely adopted and are primarily used by experts with deep knowledge about underlying architecture and compiler intricacies. This article describes the requirements that we believe necessary for making automatic performance tuning more broadly used, especially in complex, long-lived high-performance computing applications. Besides discussing limitations of current systems and strategies to overcome these, we describe the design of a system that is able to semi-automatically generate efficient platform-specific code. In the proposed system, the code optimization is programmer-guided, separately from application code, on an external file in what we call optimization programming. The language to program the optimization process is able to represent complex collections of transformations and, as a result, generate efficient platform-specific code. A database manages different optimized versions of code regions, providing a pragmatic approach to performance portability, and the framework itself has separate components, allowing the optimized code to be used on systems without installing all of the modules required for the code generation. We present experiments on two different platforms to illustrate the generation of efficient platform-specific code that performs comparable to hand-optimized, vendor-provided code.

Computer Science↗

Integrated Sensing & Controls for Coal Gasification - Development of Model-Based Controls for GE's Gasifier & Syngas Cooler. Topical Rerport for Phase III

This Topical Report for the final Phase III of the program summarizes the results from the Task 3 of the program. In this task, the separately designed extended Kalman Filter (EKF) and model predictive controls (MPC) with ideal sensing, developed in Phase II, were integrated to achieve the overall sensing and control system for the gasification section of an IGCC plant. The EKF and MPC algorithms were updated and re-tuned to achieve closed-loop system stability as well as good steady-state and transient control response. In particular, the performance of the integrated EKF and MPC solution was tested extensively through multiple simulation studies to achieve improved steady-state as well as transient performance, with coal as well as coal-petcoke blended fuel, in the presence of unknown modeling errors as well as sensor errors (noise and bias). The simulation studies demonstrated significant improvements in steady state and transient operation performance, similar to that achieved by MPC with ideal sensors in Phase II of the program.

20 FOSSIL-FUELED POWER PLANTS↗

Balancing Power Absorption Against Structural Loads With Viscous Drag and Power-Takeoff Efficiency Considerations

The focus of this paper is to balance power absorption against structural loading for a novel fixed-bottom oscillating surge wave energy converter in both regular and irregular wave environments. The power-to-load ratio will be evaluated using pseudospectral control (PSC) to determine the optimum power-takeoff (PTO) torque based on a multiterm objective function. This paper extends the pseudospectral optimal control problem to not just maximize the time-averaged absorbed power but also include measures for the surge-foundation force and PTO torque in the optimization. The objective function may now potentially include three competing terms that the optimizer must balance. Separate weighting factors are attached to the surge-foundation force and PTO control torque that can be used to tune the optimizer performance to emphasize either power absorption or load shedding. To correct the pitch equation of motion, derived from linear hydrodynamic theory, a quadratic-viscous-drag torque has been included in the system dynamics; however, to continue the use of quadratic programming solvers, an iteratively obtained linearized drag coefficient was utilized that provided good accuracy in the predicted pitch motion. Furthermore, the analysis considers the use of a nonideal PTO unit to more accurately evaluate controller performance. The PTO efficiency is not directly included in the objective function but rather the weighting factors are utilized to limit the PTO torque amplitudes, thereby reducing the losses resulting from the bidirectional energy flow through a nonideal PTO. Furthermore, results from PSC show that shedding a portion of the available wave energy can lead to greater reductions in structural loads, peak-to-average power ratio, and reactive power requirement.

16 TIDAL AND WAVE POWER↗

Portfolio-Scale Optimization of Customer Energy Efficiency Incentive and Marketing: Cooperative Research and Development Final Report, CRADA Number CRD-13-535

North East utility National Grid (NGrid) is developing a portfolio-scale application of OpenStudio designed to optimize incentive and marketing expenditures for their energy efficiency (EE) programs. NGrid wishes to leverage a combination of geographic information systems (GIS), public records, customer data, and content from the Building Component Library (BCL) to form a JavaScript Object Notation (JSON) input file that is consumed by an OpenStudio-based expert system for automated model generation. A baseline model for each customer building will be automatically tuned using electricity and gas consumption data, and a set of energy conservation measures (ECMs) associated with each NGrid incentive program will be applied to the model. The simulated energy performance and return on investment (ROI) will be compared with customer hurdle rates and available incentives to A) optimize the incentive required to overcome the customer hurdle rate and B) determine if marketing activity associated with the specific ECM is warranted for that particular customer. Repeated across their portfolio, this process will enable NGrid to substantially optimize their marketing and incentive expenditures, targeting those customers that will likely adopt and benefit from specific EE programs.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Multi-Node Program Fuzzing on High Performance Computing Resources

Significant effort is placed on tuning the internal parameters of fuzzers to explore the state space, measured as coverage, of binaries. In this work, we investigate the effects of the external environment on the resulting coverage after fuzzing two binaries with AFL for 24 hours. Parameters such as scaling to multiple nodes, node saturation, and parallel file system type on HPC resources are controlled in order to maximize coverage. It will be shown that employing a parallel file system such as IBM's General Parallel File System offers an advantage for fuzzing operations, since it contains enhancements for performance optimization. When combined with scaling to two and four nodes, while simultaneously restricting the number of coordinated AFL tasks per node on the low end (10-50% of available physical cores), coverage may be enhanced within a shorter period of time. Thus, controlling the external environment is a useful effort.

97 MATHEMATICS AND COMPUTING↗

Wide-frequency range, dynamic matching network and power system for the “Shoelace” radio frequency antenna on the Alcator C-Mod tokamak

A wide-frequency range (50–300 kHz) power system has been implemented for use with a new RF antenna – the “Shoelace” antenna – built to drive coherent plasma fluctuations in the edge of the Alcator C-Mod tokamak. A custom, dynamically tunable matching network allows two commercial 1 kW, 50-Ω RF amplifiers to drive the low-impedance, inductive load presented by the antenna. This is accomplished by a discretely variable L-match network, with 81 independently selected steps available for each of the series and parallel legs of the matching configuration. A compact programmable logic device provides a control system that measures the frequency with better than 1 kHz accuracy and transitions to the correct tuning state in less than 1 ms. At least 85% of source power is dissipated in the antenna across the operational frequency range, with a minimum frequency slew rate of 1 MHz/s; the best performance is achieved in the narrower band from 80 to 150 kHz which is of interest in typical experiments. The RF frequency can be run with open-loop control, following a pre-programmed analog waveform, or phase-locked to track a plasma fluctuation diagnostic signal in real time with programmable phase delay; the amplitude control is always open-loop. The control waveforms and phase delay are programmed remotely. These tools have enabled first-of-a-kind measurements of the tokamak edge plasma system response in the frequency range and at the wave number at which coherent fluctuations regulate heat and particle transport through the plasma boundary.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Advanced and Exploratory Shock Sensing Mechanisms.

Mechanical component response to shock environments must be predictable in order to ensure reliability and safety. Whether the shock input results from accidental drops during transportation to projectile impact scenarios, the system must irreversibly transition into a safe state that is incapable of triggering the component . With this critical need in mind, the 2017 Nuclear Weapons Summer Product Realization Institute (NW SPRINT) program objective sought the design of a passive shock failsafe with emphasis on additively manufactured (AM) components. Team Advanced and Exploratory (A&E) responded to the challenge by designing and delivering multiple passive shock sensing mech anisms that activate within a prescribed mechanical shock threshold. These AM failsafe designs were tuned and validated using analytical and computational techniques including the shock response spectrum (SRS) and finite element analysis (FEA). After rapid prototyping, the devices experienced physical shock tests conducted on Sandia drop tables to experimentally verify performance. Keywords: Additive manufacturing, dynamic system, failsafe, finite element analysis, mechanical shock, NW SPRINT, shock respon se spectrum

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗

FOCAL Campaign IV: Integrated System Control: Turbine + Hull

Campaign IV of the Floating Offshore-wind Controls Advanced Laboratory Experimental Program (FOCAL) aimed to generate a dataset enabling the validation of a floating offshore wind turbine with turbine and hull controls. The turbine considered in this campaign is the scaled IEA 15MW Reference Wind Turbine similar to the setup considered in FOCAL Campaign I. This turbine is mounted on the VolturnUS platform with integrated hull controls similar to the hull considered in FOCAL Campaign II/III. This system is deployed for a fully-coupled wind and wave testing campaign at the University of Maine's Harold Alfond Wind and Wave (W2) facility. System dynamics and turbine characteristics are measured to determine global performance and assess the effect of the control strategies employed. NREL's reference open-source controller (ROSCO) is used for active blade pitch control and hull structural control is examined through the use of tuned mass dampers (TMD's) tuned to the systems pitch and tower-bending natural frequency. The naming convention of the load cases considered are described in the data details section. Detailed properties on the modeled system are found in the following reference: Lenfest E., Floating Offshore-wind Controls Advanced Laboratory (FOCAL) Experimental Program - Campaign IV: 1:70 Model-scale Testing of the IEA-Wind 15MW Reference Turbine and the VolturnUS-S Platform. UMaine ASCC Report Number 23-57-1183.

17 WIND ENERGY↗

SU-E-J-60: Efficient Monte Carlo Dose Calculation On CPU-GPU Heterogeneous Systems

Purpose: It is well-known that the performance of GPU-based Monte Carlo dose calculation implementations is bounded by memory bandwidth. One major cause of this bottleneck is the random memory writing patterns in dose deposition, which leads to several memory efficiency issues on GPU such as un-coalesced writing and atomic operations. We propose a new method to alleviate such issues on CPU-GPU heterogeneous systems, which achieves overall performance improvement for Monte Carlo dose calculation. Methods: Dose deposition is to accumulate dose into the voxels of a dose volume along the trajectories of radiation rays. Our idea is to partition this procedure into the following three steps, which are fine-tuned for CPU or GPU: (1) each GPU thread writes dose results with location information to a buffer on GPU memory, which achieves fully-coalesced and atomic-free memory transactions; (2) the dose results in the buffer are transferred to CPU memory; (3) the dose volume is constructed from the dose buffer on CPU. We organize the processing of all radiation rays into streams. Since the steps within a stream use different hardware resources (i.e., GPU, DMA, CPU), we can overlap the execution of these steps for different streams by pipelining. Results: We evaluated our method using a Monte Carlo Convolution Superposition (MCCS) program and tested our implementation for various clinical cases on a heterogeneous system containing an Intel i7 quad-core CPU and an NVIDIA TITAN GPU. Comparing with a straightforward MCCS implementation on the same system (using both CPU and GPU for radiation ray tracing), our method gained 2-5X speedup without losing dose calculation accuracy. Conclusion: The results show that our new method improves the effective memory bandwidth and overall performance for MCCS on the CPU-GPU systems. Our proposed method can also be applied to accelerate other Monte Carlo dose calculation approaches. This research was supported in part by NSF under Grants CCF-1217906, and also in part by a research contract from the Sandia National Laboratories.

60 APPLIED LIFE SCIENCES↗

Highlights and Lessons Learned from Army Re-tuning™ Pilot Demonstrations

The Army has aggressively pursued policy and energy efficiency projects and programs in recent years to meet Federal mandates and reduction goals organizationally (EISA 2007, AR 420-1, 42 U.S.C. § 8253), resulting in a 9.6% reduction in energy use intensity (EUI) from FY2015 to FY2017. Despite these successes, the Army remains the largest consumer of electricity in the Federal government and spent approximately $1.1 billion in energy-related costs in FY2017. Many of the low-cost/no-cost conservation measures have been deployed, so finding additional savings and projects that either (a) are cost effective or (b) can be bundled with more costly capital improvement projects to improve life cycle cost effectiveness, is imperative for the Army. Leveraging building control systems (BCS) and optimizing building performance through re-tuning represents a great opportunity for the Army. In 2016, the Pacific Northwest National Laboratory (PNNL), in support of the Assistant Secretary of the Army (Installations, Energy and Environment), i.e. ASA (IEE), was tasked in 2016 with a multi-year study and pilot demonstration to develop a business case for the potential energy and cost reduction benefits from re-tuning efforts for the Army.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Final Scientific/Technical Report-FREE-piston Zero Emissions Refrigerator (FREEZER)

The FREE-piston Zero Emissions Refrigerator (FREEZER™) is a natural refrigerant Stirling-cycle cooler developed by AMSC in partnership with Heatcraft, targeted as a replacement refrigeration system for walk-in coolers or freezers found commonly in restaurants, convenience stores, supermarkets, etc. The FREEZER™, combined with a closed-loop CO2 heat-transfer system, had the goal to provides an efficient (Target COP = 1.4), zero ODP (Ozone Depletion Potential) and near zero GWP (Global Warming Potential) refrigeration solution with no synthetic, flammable, or toxic refrigerants. Other near zero GWP refrigeration systems do not exist at the scale and performance level of FREEZER. ARPA-E Continuation Funding enabled the design and fabrication of one Advanced Development Unit (ADU) cooler and two Field Test Unit (FTU) coolers, and a new revised Field Test Unit (new FTU). Heatcraft was to extensively test the various coolers at their facilities, which closely simulate real-world conditions. The original plan was to install and operate FREEZER FTUs alongside a traditional cooling solution at one or more customer locations to obtain data on direct performance comparisons with existing commercial refrigeration systems. Success in this FREEZER refinement and field-test project will establish the basis for a new class of environmentally friendly refrigeration systems that will revolutionize the industry. The hardware developed under the program is shown below in Figure 1. Five Stirling-cycle machines were built within the program, with a sixth cooler still under construction. The original DU and the ADU coolers met the performance target with a coefficient of performance (COP) of 1.2. The FTU was underperforming during the initial tuning of the free-piston Stirling cycle dynamics when there was a catastrophic failure of the fasteners on pressure boundary. This caused a pause to the program while the failure mechanism was investigated. The root cause was an inadequate flange design that allowed an excessive cyclic loading on the fasteners, leading to a fatigue failure. It was determined that existing FTU design could not be reasonably salvaged in a safe manner, and that fact, combined with the lower-than-expected performance of the new experimental heat exchanger modules lead to the design of the “new FTU” with a more conventional shell-and-tube heat exchanger design which should get close to the design target COP of 1.4, but have a higher overall machine cost. The program ran out of time before the fabrication of the new FTU was completed, but AMSC is still working on the cooler, with the majority of the parts fabricated, and hopes to have the unit under test sometime in early 2022, after which it will be shipped to Heatcraft for additional testing.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Towards Achieving Performance Portability Using Directives for Accelerators

In this paper we explore the performance portability of directives provided by OpenMP 4 and OpenACC to program various types of node architectures with attached accelerators, both self-hosted multicore and offload multicore/GPU. Our goal is to examine how successful OpenACC and the newer offload features of OpenMP 4.5 are for moving codes between architectures, how much tuning might be required and what lessons we can learn from this experience. To do this, we use examples of algorithms with varying computational intensities for our evaluation, as both compute and data access efficiency are important considerations for overall application performance. We implement these kernels using various methods provided by newer OpenACC and OpenMP implementations, and we evaluate their performance on various platforms including both X86_64 with attached NVIDIA GPUs, self-hosted Intel Xeon Phi KNL, as well as an X86_64 host system with Intel Xeon Phi coprocessors. In this paper, we explain what factors affected the performance portability such as how to pick the right programming model, its programming style, its availability on different platforms, and how well compilers can optimize and target to multiple platforms.

97 MATHEMATICS AND COMPUTING↗

Effective Source-to-Source Outlining to Support Whole Program Empirical Optimization

Although automated empirical performance optimization and tuning is well-studied for kernels and domain-specific libraries, a current research grand challenge is how to extend these methodologies and tools to significantly larger sequential and parallel applications. In this context, we present the ROSE source-to-source outliner, which addresses the problem of extracting tunable kernels out of whole programs, thereby helping to convert the challenging whole-program tuning problem into a set of more manageable kernel tuning tasks. Our outliner aims to handle large scale C/C++, Fortran and OpenMP applications. A set of program analysis and transformation techniques are utilized to enhance the portability, scalability, and interoperability of source-to-source outlining. More importantly, the generated kernels preserve performance characteristics of tuning targets and can be easily handled by other tools. Preliminary evaluations have shown that the ROSE outliner serves as a key component within an end-to-end empirical optimization system and enables a wide range of sequential and parallel optimization opportunities.

99 GENERAL AND MISCELLANEOUS↗

Choosing RF system for low energy RHIC operation

While becoming an important part of RHIC physics program, the low energy RHIC operations present substantial challenges. There is a significant body of both theoretical and experimental studies addressing these problems and possible remedies. While main physics challenge in reaching high luminosity rests with the space charge induced tune spread, there are other technical limitations strongly affecting the luminosity, life-time and overall RHIC performance in low energy collisions.

43 PARTICLE ACCELERATORS↗

Compare Energy Use in Variable Refrigerant Flow Heat Pumps Field Demonstration and Computer Model

Variable Refrigerant Flow (VRF) heat pumps are often regarded as energy efficient air-conditioning systems which offer electricity savings as well as reduction in peak electric demand while providing improved individual zone setpoint control. One of the key advantages of VRF systems is minimal duct losses which provide significant reduction in energy use and duct space. However, there is limited data available to show their actual performance in the field. Since VRF systems are increasingly gaining market share in the US, it is highly desirable to have more actual field performance data of these systems. An effort was made in this direction to monitor VRF system performance over an extended period of time in a US national lab test facility. Due to increasing demand by the energy modeling community, an empirical model to simulate VRF systems was implemented in the building simulation program EnergyPlus. This paper presents the comparison of energy consumption as measured in the national lab and as predicted by the program. For increased accuracy in the comparison, a customized weather file was created by using measured outdoor temperature and relative humidity at the test facility. Other inputs to the model included building construction, VRF system model based on lab measured performance, occupancy of the building, lighting/plug loads, and thermostat set-points etc. Infiltration model inputs were adjusted in the beginning to tune the computer model and then subsequent field measurements were compared to the simulation results. Differences between the computer model results and actual field measurements are discussed. The computer generated VRF performance closely resembled the field measurements.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

C3F: Collaborative Container-based Model Coupling Framework

Solving complex real-world grand challenge problems requires in-depth collaboration of researchers from multiple disciplines. Such collaboration often involves harnessing multiscale and multi-dimensional data and combining models from different fields to simulate systems. However, the progress on this front has been limited mainly due to significant gaps in domain knowledge and tools that are typically employed in silos of the domains. Researchers from different fields face considerable barriers to understanding and reusing each other’s data/models in order to collaborate effectively. For example, in solving the global sustainability problems, researchers from hydrology, climate science, agriculture, and economics need to run their respective models to study different components of the global and local food, energy and water systems while, at the same time, need to interact with other researchers and integrate the results of one model with another. Developing this kind of model coupling workflow calls for (1) a large amount of data being processed and exchanged across domains and organizations, (2) identifying and processing the output of one model to make it ready for integration into another model, (3) controlling the workflow dynamically so that it runs until a certain convergence condition or other criteria is met, and (4) close collaboration among the modelers to explore, tune, and test the configuration and data transformation needed to link the models. We have developed C3F, a flexible collaborative model coupling framework to help researchers accelerate their model integration and linking efforts by leveraging advanced cyberinfrastructure such as high-performance computing and virtual containers. In this paper, we describe our experience and lessons learned in developing this cyberinfrastructure solution to support the linking of Water Balance Model (WBM) and SIMPLE-G agricultural economic model in an NSF funded INFEWS project and a DOE-funded Program on Coupled Human and Earth Systems (PCHES) to study the implications of groundwater scarcity for food-energy-water systems. The C3F model coupling framework can be extended to facilitate other model linkages as well.

containerization↗

Tartan: Evaluating Modern GPU Interconnect via a Multi-GPU Benchmark Suite

In this paper, we fill the gap by proposing a multi-GPU benchmark suite named Tartan, which contains microbenchmarks, scale-up and scale-out applications. We then apply Tartan to evaluate the four latest types of modern GPU interconnects, i.e., PCI- e, NVLink-V1, NVLink-V2 and InfiniBand with GPUDirect- RDMA from two recently released NVIDIA super AI platforms as well as ORNL’s exascale prototype system. Based on empirical evaluation, we observe four new types of NUMA effects: three types are triggered by NVLink’s topology, connectivity and routing, while one type is caused by PCI-e (i.e., anti-locality). They are very important for performance tuning in multi-GPU environment. Our evaluation results show that, unless the current CPU-GPU master-slave programming model can be replaced, it is difficult for scale-up multi-GPU applications to really benefit from faster intra-node interconnects such as NVLinks; while for inter-node scale-out applications, although interconnect is more crucial to the overall performance, GPUDirect-RDMA appears to be not always the optimal choice.

HPC↗