Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “workload”

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 955 records · Page 53

Business Case Analysis for Artificial Intelligence-Large Language Model Technology Integration

AI-assisted processes are expected to enhance operational efficiency and improve decision-making, supporting the long-term economic viability of nuclear power plants. However, detailed business analyses of AI-generated cost savings are rarely performed. Given the recent industry interest in Large Language Model (LLM), the U.S. Department of Energy (DOE) Light Water Reactor Sustainability (LWRS) Program has conducted a comprehensive business case analysis of LLM Artificial Intelligence (AI) implementation in nuclear plant engineering workflows. The research employed three complementary business case approaches to evaluate impact of an LLM, using three representative engineering processes as use-cases: Boric Acid Corrosion (BAC) Evaluations, Maintenance Rule Evaluations, and 10 CFR 50.59 Screenings. Through detailed workload analyses and structured interviews, the study quantified significant efficiency improvements ranging from 11% to 59% across these processes. The research further considers how these efficiency gains could translate into tangible reliability improvements through enhanced engineering capacity. Analysis of historical plant trip data indicates that enabling engineers to focus on proactive reliability activities could provide substantial financial benefits through avoided outages, potentially generating greater value than the direct efficiency improvements alone. By documenting successful applications, implementation challenges, and strategic opportunities, this research provides nuclear utilities with a practical framework for evaluating the value of AI technology to support long-term operations through advanced digital technologies.

97 MATHEMATICS AND COMPUTING↗

Data Center Cybersecurity, Supply Chain Risk Management, and Emerging Regulation Cohort Summary: Takeaways and Action Plans

This report summarizes the outcomes of the Data Center Cohort under the Department of Energy’s Technical Assistance for Digital Assurance (TADA) initiative, aimed at enhancing grid resilience through cybersecurity, supply chain risk management (SCRM), and Cyber-Informed Engineering (CIE). The cohort engaged 17 organizations across utilities, data center operators, vendors, and technology providers in three sessions combining presentations, discussions, and exercises. Key topics included AI-driven load behavior, cybersecurity vulnerabilities in UPS/BESS and cooling systems, governance gaps at utility–data center boundaries, and supply chain integrity. Five cross-cutting themes emerged: interconnection architecture vulnerabilities, fragmented governance, AI-driven stability risks, lack of regulatory frameworks, and long-term supply chain concerns. Actionable recommendations were developed, including implementing DMZ segmentation, formalizing vendor access agreements, designing AI workload limits, and advancing standards through NERC and state-level programs. These strategies aim to strengthen resilience, clarify responsibilities, and ensure secure integration of data centers into the grid.

24 - POWER TRANSMISSION AND DISTRIBUTION↗

Operational Evolution of FTS3: A DevOps Driven Approach to Elastic Operations

The File Transfer Service (FTS3) is a distributed data movement service developed at CERN and widely used to transfer data across the Worldwide LHC Computing Grid (WLCG). At Fermilab, FTS3 supports data transfers for multiple experiments, including Intensity Frontier experiments such as DUNE, enabling reliable data movement between WebDAV endpoints in Europe and the Americas.​ At CHEP 2021, we reported on the initial containerized deployment of FTS3 on OKD, the community Kubernetes distribution of Red Hat OpenShift. In this work, we present the subsequent evolution of this deployment, focusing on new operational capabilities introduced to improve scalability, robustness, and long-term maintainability.​ We describe the adoption of more secure and reproducible container build workflows, the integration of DevOps-driven operational practices, and enhancements in monitoring and automation. A key new result is the introduction of horizontal scaling and elastic resource management, allowing FTS3 components to dynamically adapt to workload variations while maintaining service reliability. We also discuss improvements in fault tolerance and operational procedures derived from production experience.​ Finally, we summarize lessons learned from operating FTS3 as a Kubernetes-native service and outline how these developments have improved the resilience and efficiency of data movement operations at Fermilab.

Munoz Flores, Victor Leopoldo [Fermilab]↗

Modernizing GlideinWMS Factory Monitoring with Prometheus & Grafana

Large-scale scientific experiments like CMS and DUNE rely on the distributed workload management system GlideinWMS to efficiently utilize computing resources across heterogeneous computing environments. GlideinWMS currently records Factory statistics using Round Robin Databases (RRDBs), XML, and JSON files, and these statistics are displayed via custom monitoring Web pages, thereby limiting integration with modern observability platforms. This project investigates the use of Prometheus-based instrumentation to expose Factory metrics using OpenTelemetry principles. Factory statistics related to Glidein submission and job execution are exported as Prometheus metrics through the Prometheus Python Client Library and are served via an HTTP metrics endpoint. The collected metrics are inspected using the Prometheus web-based interface and are visualized through Grafana dashboards within the Landscape monitoring infrastructure at Fermilab. This project significantly streamlines the integration of modern monitoring technologies into GlideinWMS and establishes a framework for extending observability across additional system components.

Appiah, Gideon [Grambling State U.]↗

A Methodology to Evaluate the Grid Reliability Impact of Oscillations Induced by Large Loads

The rapid growth of hyperscale AI data centers is bringing renewed attention to the reliability risk that sustained forced oscillations pose to bulk power systems, with cyclic computational workloads emerging as a new forcing source. Unlike the broadband, stochastic disturbances from traditional industrial loads such as arc furnaces, AI training and inference facilities can inject large active power swings concentrated at specific frequencies over extended durations - characteristics that existing grid planning practices do not account for. While the North American Electric Reliability Corporation (NERC) has recognized this gap and called for system-level studies of large load interconnections, no standardized methodology exists to screen, simulate, and quantify these risks at the planning stage. This report presents the Risk Assessment Tool for Large Load-induced Events (RATLLE), a Python-based, publicly available script suite developed at the Pacific Northwest National Laboratory to evaluate bulk power system reliability risks from data center-induced oscillations. RATLLE implements a three-module workflow: a screening module that identifies vulnerable interconnection locations and excitable system modes; a simulation module that models cyclic data center load behavior using a commercial positive sequence simulation platform; and an analysis module that computes risk metrics and generates interactive visualization dashboards. The risk metrics, formulated around simulation observables, map oscillation impacts to a three-stage severity scale spanning latent equipment fatigue through imminent cascading failure. The methodology is demonstrated on two Western Electricity Coordinating Council (WECC) system models: a publicly available 240-bus reduced representation and a detailed 2031 Heavy Winter planning case. Case studies illustrate that even modest 50 MW forced oscillations at resonant frequencies can produce wide-area power swings, N-1 security constraint violations, and cascading generator trips through protection actions - outcomes that would not occur under normal operating conditions without oscillations present. The results underscore the need for standardized oscillation impact assessment in large load interconnection studies and provide a reproducible, extensible framework for utilities to adopt or customize within their existing planning workflows.

Biswas, Shuchismita↗

Parallel job scheduling policies to improve fairness : a case study.

Balancing fairness, user performance, and system performance is a critical concern when developing and installing parallel schedulers. Sandia uses a customized scheduler to manage many of their parallel machines. A primary function of the scheduler is to ensure that the machines have good utilization and that users are treated in a 'fair' manner. A separate compute process allocator (CPA) ensures that the jobs on the machines are not too fragmented in order to maximize throughput. Until recently, there has been no established technique to measure the fairness of parallel job schedulers. This paper introduces a 'hybrid' fairness metric that is similar to recently proposed metrics. The metric uses the Sandia version of a 'fairshare' queuing priority as the basis for fairness. The hybrid fairness metric is used to evaluate a Sandia workload. Using these results, multiple scheduling strategies are introduced to improve performance while satisfying user and system performance constraints.

99 GENERAL AND MISCELLANEOUS↗

A File Allocation Strategy for Energy-Efficient Disk Storage Systems

Exponential data growth is a reality for most enterprise and scientific data centers.Improvements in price/performance and storage densities of disks have made it both easy and affordable to maintain most of the data in large disk storage farms. The provisioning of disk storage farms however, is at the expense of high energy consumption due to the large number of spinning disks. The power for spinning the disks and the associated cooling costs is a significant fraction of the total power consumption of a typical data center. Given the trend of rising global fuel and energy prices and the high rate of data growth, the challenge is to implement appropriateconfigurations of large scale disk storage systems that meet performancerequirements for information retrieval across data centers. We present part of the solution to this challenge with an energy efficient file allocation strategy on a large scale disk storage system. Given performance characteristics of thedisks, and a profile of the workload in terms of frequencies of file requests and their sizes, the basic idea is to allocate files to disks such that the disks can be configured into two sets of active (constantly spinning), and passive (capable of being spun up or down) disk pools. The goal is to minimize the number of active disks subject to I/O performance constraints. We present an algorithm for solving this problem with guaranteed bounds from the optimal solution. Our algorithm runs in O(n) time where n is the number of files allocated. It uses a mapping of our file allocation problem to a generalization of the bin packing problem known as 2-dimensional vector packing. Detailed simulation results are also provided.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Measuring FLOPS Using Hardware Performance Counter Technologies on LC systems

FLOPS (FLoating-point Operations Per Second) is a commonly used performance metric for scientific programs that rely heavily on floating-point (FP) calculations. The metric is based on the number of FP operations rather than instructions, thereby facilitating a fair comparison between different machines. A well-known use of this metric is the LINPACK benchmark that is used to generate the Top500 list. It measures how fast a computer solves a dense N by N system of linear equations Ax=b, which requires a known number of FP operations, and reports the result in millions of FP operations per second (MFLOPS). While running a benchmark with known FP workloads can provide insightful information about the efficiency of a machine's FP pipelines in relation to other machines, measuring FLOPS of an arbitrary scientific application in a platform-independent manner is nontrivial. The goal of this paper is twofold. First, we explore the FP microarchitectures of key processors that are underpinning the LC machines. Second, we present the hardware performance monitoring counter-based measurement techniques that a user can use to get the native FLOPS of his or her program, which are practical solutions readily available on LC platforms. By nature, however, these native FLOPS metrics are not directly comparable across different machines mainly because FP operations are not consistent across microarchitectures. Thus, the first goal of this paper represents the base reference by which a user can interpret the measured FLOPS more judiciously.

99 GENERAL AND MISCELLANEOUS↗

Certification of Completion of ASC FY08 Level-2 Milestone ID #2933

This report documents the satisfaction of the completion criteria associated with ASC FY08 Milestone ID No.2933: 'Deploy Moab resource management services on BlueGene/L'. Specifically, this milestone represents LLNL efforts to enhance both SLURM and Moab to extend Moab's capabilities to schedule and manage BlueGene/L, and increases portability of user scripts between ASC systems. The completion criteria for the milestone are the following: (1) Batch jobs can be specified, submitted to Moab, scheduled and run on the BlueGene/L system; (2) Moab will be able to support the markedly increased scale in node count as well as the wiring geometry that is unique to BlueGene/L; and (3) Moab will also prepare and report statistics of job CPU usage just as it does for the current systems it supports. This document presents the completion evidence for both of the stated milestone certification methods: Completion evidence for this milestone will be in the form of (1) documentation--a report that certifies that the completion criteria have been met; and (2) user hand-off. As the selected Tri-Lab workload manager, Moab was chosen to replace LCRM as the enterprise-wide scheduler across Livermore Computing (LC) systems. While LCRM/SLURM successfully scheduled jobs on BG/L, the effort to replace LCRM with Moab on BG/L represented a significant challenge. Moab is a commercial product developed and sold by Cluster Resources, Inc. (CRI). Moab receives the users batch job requests and dispatches these jobs to run on a specific cluster. SLURM is an open-source resource manager whose development is managed by members of the Integrated Computational Resource Management Group (ICRMG) within the Services and Development Division at LLNL. SLURM is responsible for launching and running jobs on an individual cluster. Replacing LCRM with Moab on BG/L required substantial changes to both Moab and SLURM. While the ICRMG could directly manage the SLURM development effort, the work to enhance Moab had to be done by Moab's vendor. Members of the ICRMG held many meetings with CRI developers to develop the design and specify the requirements for what Moab needed to do. Extensions to SLURM are used to run jobs on the BlueGene/L architecture. These extensions support the three dimensional network topology unique to BG/L. While BG/L geometry support was already in SLURM, enhancements were needed to provide backfill capability and answer 'will-run' queries from Moab. For its part, the Moab architecture needed to be modified to interact with SLURM in a more coordinated way. It needed enhancements to support SLURM's shorthand notation for representing thousands of compute nodes and report this information using Moab's existing status commands. The LCRM wrapper scripts that emulated LCRM commands also needed to be enhanced to support BG/L usage. The effort was successful as Moab 5.2.2 and SLURM 1.3 was installed on the 106496 node BG/L machine on May 21, 2008, and turned over to the users to run production.

99 GENERAL AND MISCELLANEOUS↗

INFORMATION: Special Report on "Selected Department of Energy Program Efforts to Implement the American Recovery and Reinvestment Act"

The American Recovery and Reinvestment Act of 2009 (Recovery Act) was enacted on February 17, 2009, to jumpstart the economy by creating or saving millions of jobs, spurring technological advances in health and science, and investing in the Nation's energy future. The Department of Energy received over $32.7 billion in Recovery Act funding for various science, energy, and environmental programs and initiatives. As of November 2009, the Department had obligated $18.3 billion of the Recovery Act funding, but only $1.4 billion had been spent. The Department's Offices of Energy Efficiency and Renewable Energy, Fossil Energy, Environmental Management, Science, and Electricity Delivery and Energy Reliability received the majority of funding allocated to the Department, about $32.3 billion. Obligating these funds by the end of Fiscal Year 2010, as required by the Recovery Act, and overseeing their effective use in succeeding years, represents a massive workload increase for the Department's programs. The effort to date has strained existing resources. As has been widely acknowledged, any effort to disburse massive additional funding and to expeditiously initiate and complete projects increases the risk of fraud, waste and abuse. It is, therefore, important for the Department's program offices to assess and mitigate these risks to the maximum extent practicable. In this light, we initiated this review as an initial step in the Office of Inspector General's charge to determine whether the Department's major program offices had developed an effective approach for identifying and mitigating risks related to achieving the goals and objectives of the Recovery Act. The Department's program offices included in our review identified risks and planned mitigation strategies that, if successfully implemented and executed, should help achieve the goals and objectives of the Recovery Act. While each office identified risks unique to its respective areas of responsibility, there were a number of risks shared in common. These included the mechanical and substantive requirements related to the award and distribution of funds; program and project performance monitoring; and, program and project execution activities. In particular, the offices self-identified common risks such as: (1) The inability to award and distribute funds in a timely manner to achieve the goals of the Recovery Act; (2) The sufficiency of monitoring procedures and resources to, among other things, prevent and detect fraud, waste and abuse throughout the performance period of financial assistance awards and contracts; and (3) The inherent cost, schedule and performance risks associated with first-of-a-kind, innovative research and demonstration projects. Our review confirmed that the Department had begun to implement a number of strategies designed to mitigate these and other program-specific risks. Our testing, however, identified challenges to the effective implementation of these mitigation strategies that need to be addressed if the Department is to meet the goals and objectives established by the Recovery Act. At the time of our review: (1) Program staffing resources, critical to the success of all other mitigation strategies, remained inadequate both in numbers and qualifications (certifications and training) for positions in procurement and acquisition, project management, and monitoring and oversight functions; (2) Performance measures for achieving Recovery Act goals such as distributing funds in an expeditious manner had not always been established and included in performance plans, and, in financial assistance and contract documents; and (3) Programs had not consistently demonstrated that previously reported deficiencies, identified through audits, inspections, investigations and other oversight activities, had been considered in designing mitigation strategies for the Recovery Act related risks. As we noted in our Special Report on the American Recovery and Reinvestment Act at the Department of Energy (OAS-RA-09-01, March 2009) these sorts of deficiencies, which were described in the March 2009 report, had adversely impacted the success of earlier Departmental projects. We concluded that consideration of these previously encountered internal control weaknesses was essential if the Department was to avoid the same or similar problems in executing the Recovery Act programs/projects. During the course of our audit we identified other internal control weakness indicators. When brought to their attention by the Office of Inspector General audit staff, program officials acted promptly to address these problems by modifying or improving their risk mitigation strategies. For instance, we found that the Office of Energy Efficiency and Renewable Energy, responsible for $16.8 billion in Recovery Act program activities, took steps to strengthen its merit review process to avoid potential conflict of interest concerns.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Preliminary analysis of the International Data Centre pipeline.

The International Data Centre of the Comprehensive Nuclear-Test-Ban Treaty Organization relies on automatic data processing as the first step in identifying seismic events from seismic waveform data. However, more than half of the automatically identified seismic events are eliminated by IDC analysts. Here, an IDC dataset is analyzed to determine if the number of automatically generated false positives could be reduced. Data that could be used to distinguish false positives from analyst-accepted seismic events includes the number of stations, the number of phases, the signal-to-noise ratio, and the pick error. An empirical method is devised to determine whether an automatically identified seismic event is acceptable, and the method is found to identify a significant number of the false positives in IDC data. This work could help reduce seismic analyst workload and could help improve the calibration of seismic monitoring stations. This work could also be extended to address identification of seismic events missed by automatic processing.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗

WILD PIGS: BIOLOGY, DAMAGE, CONTROL TECHINQUES AND MANAGEMENT

The existence of problems with wild pigs (Sus scrofa) is nothing new to the Western Hemisphere. Damage by these introduced animals was reported as far back as 1505 by the early Spanish colonies in the Caribbean, where wild pigs were killing the colonists cattle. Droves of these animals also ravaged cultivated crops of maize and sugarcane on islands in the West Indies during this same time period. These wild pigs reportedly were very aggressive and often attacked Spanish soldiers hunting rebellious Indians or escaped slaves on these islands, especially when these animals were cornered. The documentation of such impacts by introduced populations of this species in the United States has subsequently increased in recent years, and continued up through the present (Towne and Wentworth. 1950, Wood and Barrett 1979, Mayer and Brisbin 1991, Dickson et al. 2001). In spite of a fairly constant history in this country since the early 1900s, wild pigs have had a dramatic recent increase in both distribution and numbers in the United States. Between 1989 and 2009, the number of states reporting the presence of introduced wild pigs went from 19 up to as many as 44. This increase, in part natural, but largely manmade, has caused an increased workload and cost for land and resource managers in areas where these new populations are found. This is the direct result of the damage that these introduced animals do. The cost of both these impacts and control efforts has been estimated to exceed a billion dollars annually (Pimentel 2007). The complexity of this problem has been further complicated by the widespread appeal and economic potential of these animals as a big game species (Tisdell 1982, Degner 1989). Wild pigs are a controversial problem that is not going away and will likely only get worse with time. Not only do they cause damage, but wild pigs are also survivors. They reproduce at a rate faster than any other mammal of comparable size, native or introduced; they can eat just about anything; and, they can live just about anywhere. On top of that, wild pigs are both very difficult to control and, with the possible exception of island ecosystems, almost impossible to eradicate (Dickson et al. 2001, Sweeney et al. 2003). The solution to the wild pig problem has not been readily apparent. The ultimate answer as to how to control these animals has not been found to date. In many ways, wild pigs are America's most successful large invasive species. All of which means that wild pigs are a veritable nightmare for land and resource managers trying to keep the numbers of these animals and the damage that they do under control. Since the more that one knows about an invasive species, the easier it is to deal with and hopefully control. For wild pigs then, it is better to 'know thy enemy' than to not, especially if one expects to be able to successfully control them. In an effort to better 'know thy enemy,' a two-day symposium was held in Augusta, Georgia, on April 21-22, 2004. This symposium was organized and sponsored by U.S.D.A. Forest Service-Savannah River (USFS-SR), U. S. Department of Energy-Savannah River Operations Office (DOE-SR), the Westinghouse Savannah River Company (WSRC), the South Carolina Chapter of the Soil & Water Conservation Society, and the Savannah River Ecology Laboratory (SREL). The goal of this symposium was to assemble researchers and land managers to first address various aspects of the biology and damage of wild pigs, and then review the control techniques and management of this invasive species. The result would then be a collected synopsis of what is known about wild pigs in the United States. Although the focus of the symposium was primarily directed toward federal agencies, presenters also included professionals from academic institutions, and private-sector control contractors and land managers. Most of the organizations associated with implementing this symposium were affiliated with the Savannah River Site (SRS), a 803 km{sup 2} federal nuclear facility, located in western South Carolina along the Savannah River. The SRS was a very appropriate facility to host this symposium. The SRS has been dealing with its wild pig problem since the early 1950s. A lot has been learned about these animals at the site over the ensuing decades. Between the USFS-SR, DOE-SR, SREL, and the site's management and operations contractor, which is currently Savannah River Nuclear Solutions, SRS organizations have conducted a wealth of research on this wild pig population, spanning a broad spectrum of topics and disciplines. In fact, the SRS wild pig population is among the most studied, and possibly is the best studied population of this invasive species found in the United States today. Unfortunately, with all of that work, the ultimate answer to controlling wild pigs and their impacts still has not been found. This volume represents the collected synopsis that was the goal of the aforementioned symposium.

54 ENVIRONMENTAL SCIENCES↗

Using wesBench to Study the Rendering Performance of Graphics Processing Units

Graphics operations consist of two broad operations. The first, which we refer to here as vertex operations, consists of transformation, lighting, primitive assembly, and so forth. The second, which we refer to as pixel or fragment operations, consist of rasterization, texturing, scissoring, blending, and fill. Overall GPU rendering performance is a function of throughput of both these interdependent stages: if one stage is slower than the other, the faster stage will be forced to run more slowly and overall rendering performance will be adversely affected. This relationship is commutative: if the later stage has a greater workload than the earlier stage, the earlier stage will be forced to 'slow down.' For example, a large triangle that covers many screen pixels will incur a very small amount of work in the vertex stage while at the same time incurring a relatively large amount of work in the fragment stage. Rendering performance of a scene consisting of many large-area triangles will be limited by throughput of the fragment stage, which will have relatively more work than the vertex stage. There are two main objectives for this document. First, we introduce a new graphics benchmark, wesBench, which is useful for measuring performance of both stages of the rendering pipeline under varying conditions. Second, we present its methodology for measuring performance and show results of several performance measurement studies aimed at producing better understanding of GPU rendering performance characteristics and limits under varying configurations. First, in Section 2, we explore the 'crossover' point between geometry and rasterization. Second, in Section 3, we explore additional performance characteristics, some of which are ill- or un-documented. Lastly, several appendices provide additional material concerning problems with the gfxbench benchmark, and details about the new wesBench graphics benchmark.

97 MATHEMATICS AND COMPUTING↗

Energy Proportionality for Disk Storage Using Replication

Energy saving has become a crucial concern in datacenters as several reports predict that the anticipated energy costs over a three year period will exceed hardware acquisition. In particular, saving energy for storage is of major importance as storage devices (and cooling them off) may contribute over 25 percent of the total energy consumed in a datacenter. Recent work introduced the concept of energy proportionality and argued that it is a more relevant metric than just energy saving as it takes into account the tradeoff between energy consumption and performance. In this paper, we present a novel approach, called FREP (Fractional Replication for Energy Proportionality), for energy management in large datacenters. FREP includes areplication strategy and basic functions to enable flexible energy management. Specifically, our method provides performance guarantees by adaptively controlling the power states of a group of disks based on observed and predicted workloads. Our experiments, using a set of real and synthetic traces, show that FREP dramatically reduces energy requirements with a minimal response time penalty.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

FPGA Acceleration of GCN in Light of the Symmetry of Graph Adjacency Matrix

Graph Convolutional Neural Networks (GCNs) are widely used to process large-scale graph data. Different from deep neural networks (DNNs), GCNs are sparse, irregular, and unstructured, posing unique challenges to hardware acceleration with regular processing elements (PEs). In particular, the adjacency matrix of a GCN is extremely sparse, leading to frequent but irregular memory access, low spatial/temporal data locality and poor data reuse. Furthermore, a realistic graph usually consists of unstructured data (e.g., unbalanced distributions), creating significantly different processing times and imbalanced workload for each node in GCN acceleration. To overcome these challenges, we propose an end-to-end hardware-software co-design to accelerate GCNs on resource-constrained FPGAs with the features including: (1) A custom dataflow that leverages symmetry along the diagonal of the adjacency matrix to accelerate feature aggregation for undirected graphs. We utilize either the upper or the lower triangular matrix of the adjacency matrix to perform aggregation in GCN to improve data reuse. (2) Unified compute cores for both aggregation and transform phases, with full support to the symmetry-based dataflow. These cores can be dynamically reconfigured to the systolic mode for transformation or as individual accumulators for aggregation in GCN processing. (3) Preprocessing of the graph in software to rearrange the edges and features to match the custom dataflow. This step improves the regularity in memory access and data reuse in the aggregation phase. Moreover, we quantize the GCN precision from FP32 to INT8 to reduce the memory footprint without losing the inference accuracy. We implement our accelerator design in Intel Stratix10 MX FPGA board with HBM2, and demonstrate 1.3x-110.5x improvement in end-to-end GCN latency as compared to the state-of the-art FPGA implementations, on the graph datasets of Cora, Pubmed, Citeseer and Reddit.

Nair, Gopikrishnan R.↗

Fine-Grained Power and Energy Attribution on AMD GPU/APU-Based Exascale Nodes

Modern exascale GPU- and APU-based systems provide multiple power and energy sensors, but differences in scope, update rate, timing, and filtering complicate the attribution of short-lived accelerator activity. This paper presents a methodology to characterize and correct these effects on Cray EX systems with AMD Instinct MI250X GPUs (Frontier) and MI300A APUs (Portage). Using controlled square-wave workloads, we quantify update intervals, delay, aliasing, and variability across up to 512 GPUs and 480 APUs with on-chip (rocm-smi/amd-smi) and off-chip Cray Power Management sensors. We reconstruct power from cumulative energy counters to achieve faster response times, validate it against on-chip, off-chip, and node-level sensors, and integrate the resulting streams into a Score-P/PAPI-based tool for time-aligned, phase-level attribution. Applied to rocHPL, rocHPL-MxP, and HPG-MxP, the method separates energy savings due to reduced runtime from changes in power. Mixed precision reduces node energy on Frontier by 79% for rocHPL-MxP and 31% for HPG-MxP, with similar trends on Portage. These results provide portable guidance for sensor validation and power-aware optimization on current and future exascale systems.

Mcdaniel, Adam [ORNL] (ORCID:000000016926028X)↗

A Methodology for Evaluating Operator Usage of Machine Learning Recommendations for Power Grid Contingency Analysis

This work presents the application of a methodology to measure domain expert trust and workload, elicit feedback, and understand the technological usability and impact when a machine learning assistant is introduced into contingency analysis for real-time power grid simulation. The goal of this framework is to rapidly collect and analyze a broad variety of human factors data in order to accelerate the development and evaluation loop for deploying machine learning applications. We describe our methodology and analysis, and we discuss insights gained from a pilot participant about the current usability state of an early technology readiness level (TRL) artificial neural network (ANN) recommender.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Toward a persistent event-streaming system for high-performance computing applications

High-performance computing (HPC) applications have traditionally relied on parallel file systems and file transfer services to manage data movement and storage. Alternative approaches have been proposed that use direct communications between application components, trading persistence and fault tolerance for speed. Event-driven architectures, as popularized in enterprise contexts, present a compelling middle ground, avoiding the performance cost and API constraints of parallel file systems while retaining persistence and offering impedance matching between application components. However, adapting streaming frameworks to HPC workloads requires addressing challenges unique to HPC systems. This paper investigates the potential for a streaming framework designed for HPC infrastructures and use cases. We introduce Mofka, a persistent event-streaming framework designed specifically for HPC environments. Mofka combines the capabilities of a traditional streaming service with optimizations tailored to the HPC context, such as support for massively multicore nodes, efficient scaling for large producer-consumer workflows, RDMA-enabled high-performance network communications, specialized network fabrics with multiple links per node, and efficient handling of large scientific data payloads. Built using the Mochi suite of HPC data service components, Mofka provides a lightweight, modular, and high-performance solution for persistent streaming in HPC systems. We present the architecture of Mofka and evaluate its performance against Kafka and Redpanda using benchmarks on diverse platforms, including Argonne's Polaris and Oak Ridge's Frontier supercomputers, showing up to 8× improvement in throughput in some scenarios. We then demonstrate its utility in several real-world applications: a tomographic reconstruction pipeline, a workflow for the discovery of metal-organic frameworks for carbon capture, and the instrumentation of Dask workflows for provenance tracking and performance analysis.

HPC↗