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 415 records · Page 23

Level-2 Milestone 9009: Flux and Rabbit Capabilities on El Capitan

This document is the milestone delivery report for the ASC 2025 L2 milestone (See Table 1) for advanced I/O capabilities for El Capitan via Flux Workload Manager support and the new I/O hardware designed for El Capitan, the Rabbit Storage System. In this document we describe the design of the Rabbit Storage System and how it is managed by Flux. We evaluate the performance and usability of Rabbit using ARES, IOR, and an AI inference workload. Overall, we find that Rabbit shows good scalability, especially in node-local storage configurations, and is more scalable than the global Lustre parallel file system.

97 MATHEMATICS AND COMPUTING↗

Benchmarking DAOS Filesystem on Aurora

We benchmark the DAOS filesystem on Argonne's Aurora supercomputer (127 nodes, 4,064 targets) using fio, IOR, mdtest, and IO500 to characterize I/O and metadata performance across the DFS API and DFuse+POSIX. Single-client fio shows POSIX bandwidth saturating at 1–2 MiB I/O sizes, with write-heavy workloads outperforming reads. Multi-node IOR shows DFS bandwidth scaling well up to ~32 tasks/node, with write latency growing faster than read latency. An 8-node IO500 evaluation shows DFS achieving ~5x higher bandwidth and ~190x higher IOPS than POSIX. Results indicate DAOS is well-suited to read-heavy workloads like AI training data loading, given appropriately sized transfers and concurrency.

George, Rebecca [College of William and Mary, Will↗

DAiSES: Dynamic Adaptivity in Support of Extreme Scale Department of Energy Project No. ER25622 Prime Contract No. DE-FG02-04ER25622 Final Report for September 15, 2004-September 14, 2008

The DAiSES project [Te04] was focused on enabling conventional operating systems, in particular, those running on extreme scale systems, to dynamically customize system resource management in order to offer applications the best possible environment in which to execute. Such dynamic adaptation allows operating systems to modify the execution environment in response to changes in workload behavior and system state. The main challenges of this project included determination of what operating system (OS) algorithms, policies, and parameters should be adapted, when to adapt them, and how to adapt them. We addressed these challenges by using a combination of static analysis and runtime monitoring and adaptation to identify a priori profitable targets of adaptation and effective heuristics that can be used to dynamically trigger adaptation. Dynamic monitoring and adaptation of the OS was provided by either kernel modifications or the use of KernInst and Kperfmon [Wm04]. Since Linux, an open source OS, was our target OS, patches submitted by kernel developers and researchers often facilitated kernel modifications. KernInst operates on unmodified commodity operating systems, i.e., Solaris and Linux; it is fine-grained, thus, there were few constraints on how the underlying OS can be modified. Dynamically adaptive functionality of operating systems, both in terms of policies and parameters, is intended to deliver the maximum attainable performance of a computational environment and meet, as best as possible, the needs of high-performance applications running on extreme scale systems, while meeting system constraints. DAiSES research endeavored to reach this goal by developing methodologies for dynamic adaptation of OS parameters and policies to manage stateful and stateless resources [Te06] and pursuing the following two objectives: (1) Development of mechanisms to dynamically sense, analyze, and adjust common performance metrics, fluctuating workload situations, and overall system environment conditions. (2) Demonstration, via Linux prototypes and experiments, of dynamic self-tuning and self-provisioning in HPC environments. From a high level, the DAiSES methodology, depicted in Figure 1, includes characterization of application resource usage patterns, identification of candidate (profitable) adaptation targets, determination of feasible adaptation ranges, definition of heuristics to trigger adaptation, design and implementation of OS monitoring, triggering, and adaptation code, and quantification of performance gains.

97 MATHEMATICS AND COMPUTING↗

Dawn Usage, Scheduling, and Governance Model

This document describes Dawn use, scheduling, and governance concerns. Users started running full-machine science runs in early April 2009 during the initial open shakedown period. Scheduling Dawn while in the Open Computing Facility (OCF) was controlled and coordinated via phone calls, emails, and a small number of controlled banks. With Dawn moving to the Secure Computing Facility (SCF) in fall of 2009, a more detailed scheduling and governance model is required. The three major objectives are: (1) Ensure Dawn resources are allocated on a program priority-driven basis; (2) Utilize Dawn resources on the job mixes for which they were intended; and (3) Minimize idle cycles through use of partitions, banks and proper job mix. The SCF workload for Dawn will be inherently different than Purple or BG/L, and therefore needs a different approach. Dawn's primary function is to permit adequate access for tri-lab code development in preparation for Sequoia, and in particular for weapons multi-physics codes in support of UQ. A second purpose is to provide time allocations for large-scale science runs and for UQ suite calculations to advance SSP program priorities. This proposed governance model will be the basis for initial time allocation of Dawn computing resources for the science and UQ workloads that merit priority on this class of resource, either because they cannot be reasonably attempted on any other resources due to size of problem, or because of the unavailability of sizable allocations on other ASC capability or capacity platforms. This proposed model intends to make the most effective use of Dawn as possible, but without being overly constrained by more formal proposal processes such as those now used for Purple CCCs.

99 GENERAL AND MISCELLANEOUS↗

Rasterization with Data-Parallel Primitives

Parallel rasterization can suffer from race conditions during fragment generation, which is traditionally addressed by using specialized hardware accessible via vendor graphics APIs. Unfortunately, graphics APIs are increasingly problematic on high-performance computers, either because they are not provided or because of concerns about dependencies with in situ visualization. In response, we present a hardware-agnostic rasterization algorithm that handles race conditions using only data-parallel primitives (DPPs), enabling efficient rendering on HPC systems without graphics API dependencies and aligning with recent efforts to deliver visualization software with DPPs. Our evaluation consists of three phases: (1) evaluating portability across different CPU and GPU architectures, (2) evaluating competitiveness with a community standard, and (3) evaluating performance across varying workloads and available parallelism. The supporting experiments run on both AMD and NVIDIA GPUs, considering data sets as large as 460 million triangles and 160 million pixels. While performance generally falls short of graphics API baselines, it achieves interactive frame rates on most workloads. As a result, we conclude our approach is a viable solution for rasterization on high-performance computers since our approach is portably performant across different architectures without the need for specialized vendor support.

Buckley, Makani [University of Oregon] (ORCID:0009↗

Addressing GPU memory limitations for Graph Neural Networks in High-Energy Physics applications

Introduction Reconstructing low-level particle tracks in neutrino physics can address some of the most fundamental questions about the universe. However, processing petabytes of raw data using deep learning techniques poses a challenging problem in the field of High Energy Physics (HEP). In the Exa.TrkX Project, an illustrative HEP application, preprocessed simulation data is fed into a state-of-art Graph Neural Network (GNN) model, accelerated by GPUs. However, limited GPU memory often leads to Out-of-Memory (OOM) exceptions during training, due to the large size of models and datasets. This problem is exacerbated when deploying models on High-Performance Computing (HPC) systems designed for large-scale applications. Methods We observe a high workload imbalance issue during GNN model training caused by the irregular sizes of input graph samples in HEP datasets, contributing to OOM exceptions. We aim to scale GNNs on HPC systems, by prioritizing workload balance in graph inputs while maintaining model accuracy. Our paper introduces diverse balancing strategies aimed at decreasing the maximum GPU memory footprint and avoiding the OOM exception, across various datasets. Results Our experiments showcase memory reduction of up to 32.14% compared to the baseline. We also demonstrate the proposed strategies can avoid OOM in application. Additionally, we create a distributed multi-GPU implementation using these samplers to demonstrate the scalability of these techniques on the HEP dataset. Discussion By assessing the performance of these strategies as data loading samplers across multiple datasets, we can gauge their effectiveness in both single-GPU and distributed environments. Our experiments, conducted on datasets of varying sizes and across multiple GPUs, broaden the applicability of our work to various GNN applications that handle input datasets with irregular graph sizes.

Lee, Claire Songhyun↗

An Integrated Framework for Memory-Centric Analysis: From Trace Collection to Co-Design

The memory wall phenomenon—where advances in processor performance significantly outpace those in memory subsystems—poses a fundamental challenge for contemporary computing systems. In memory-bound applications, memory subsystem behavior dominates performance, yet existing analysis approaches present significant limitations: detailed microarchitectural simulators require days to weeks to simulate modest workloads; hardware performance counters provide only aggregate statistics that obscure temporal and spatial access patterns; and scaled simulation approaches face challenges in capturing certain behaviors that emerge at larger scales. These limitations reflect a processor-centric design philosophy increasingly misaligned with memory-bound workloads where detailed understanding of memory access patterns, cache hierarchy interactions, and contention is critical for effective optimization. This paper presents an integrated framework for memory-centric analysis that enables effective hardware-software co-design. We describe practical trace collection techniques, including hardware-assisted processor tracing with minimal overhead and portable software-based instrumentation with statistical sampling. We present multi-perspective analysis methods that examine memory behavior from temporal, sequential, spatial, and relational viewpoints, revealing distinct optimization opportunities invisible in aggregate metrics. We detail an architectural modeling framework that uses sampled traces with temporal interpolation and confidence-based filtering to evaluate cache and memory configurations. Evaluation on representative benchmarks demonstrates that this framework achieves practical accuracy (L2 cache errors of 2.64\%, confidence-filtered L3 errors of 9.92\%, bandwidth errors of 7.33\%) while providing substantial speedup (26.8×) over cycle-accurate simulation, enabling rapid design space exploration. We demonstrate how this integrated framework enables systematic identification of both hardware optimizations (memory controller tuning, bank partitioning, NUMA configuration) and software optimizations (data layout restructuring, prefetching strategies, memory-aware scheduling). Through this comprehensive treatment of the memory-centric analysis pipeline—from trace collection through architectural modeling to co-design application—we provide researchers and practitioners with practical techniques for addressing memory bottlenecks in contemporary computing systems.

Gajaria, Dhruv Mayur↗

Failing in place for low-serviceability storage infrastructure using high-parity GPU-based RAID.

In order to provide large quantities of high-reliability disk-based storage, it has become necessary to aggregate disks into fault-tolerant groups based on the RAID methodology. Most RAID levels do provide some fault tolerance, but there are certain classes of applications that require increased levels of fault tolerance within an array. Some of these applications include embedded systems in harsh environments that have a low level of serviceability, or uninhabited data centers servicing cloud computing. When describing RAID reliability, the Mean Time To Data Loss (MTTDL) calculations will often assume that the time to replace a failed disk is relatively low, or even negligible compared to rebuild time. For platforms that are in remote areas collecting and processing data, it may be impossible to access the system to perform system maintenance for long periods. A disk may fail early in a platform's life, but not be replaceable for much longer than typical for RAID arrays. Service periods may be scheduled at intervals on the order of months, or the platform may not be serviced until the end of a mission in progress. Further, this platform may be subject to extreme conditions that can accelerate wear and tear on a disk, requiring even more protection from failures. We have created a high parity RAID implementation that uses a Graphics Processing Unit (GPU) to compute more than two blocks of parity information per stripe, allowing extra parity to eliminate or reduce the requirement for rebuilding data between service periods. While this type of controller is highly effective for RAID 6 systems, an important benefit is the ability to incorporate more parity into a RAID storage system. Such RAID levels, as yet unnamed, can tolerate the failure of three or more disks (depending on configuration) without data loss. While this RAID system certainly has applications in embedded systems running applications in the field, similar benefits can be obtained for servers that are engineered for storage density, with less regard for serviceability or maintainability. A storage brick can be designed to have a MTTDL that extends well beyond the useful lifetime of the hardware used, allowing the disk subsystem to require less service throughout the lifetime of a compute resource. This approach is similar to the Xiotech ISE. Such a design can be deliberately placed remotely (without frequent support) in order to provide colocation, or meet cost goals. For workloads where reliability is key, but conditions are sub-optimal for routine serviceability, a high-parity RAID can provide extra reliability in extraordinary situations. For example, for installations requiring very high Mean Time To Repair, the extra parity can eliminate certain problems with maintaining hot spares, increasing overall reliability. Furthermore, in situations where disk reliability is reduced because of harsh conditions, extra parity can guard against early data loss due to lowered Mean Time To Failure. If used through an iSCSI interface with a streaming workload, it is possible to gain all of these benefits without impacting performance.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Comparing Coordinated Garbage Collection Algorithms for Arrays of Solid-state Drives

Solid-State Drives (SSDs) offer significant performance improvements over hard disk drives (HDD) on a number of workloads. The frequency of garbage collection (GC) activity is directly correlated with the pattern, frequency, and volume of write requests, and scheduling of GC is controlled by logic internal to the SSD. SSDs can exhibit significant performance degradations when garbage collection (GC) conflicts with an ongoing I/O request stream. When using SSDs in a RAID array, the lack of coordination of the local GC processes amplifies these performance degradations. No RAID controller or SSD available today has the technology to overcome this limitation. In our previous work, we presented a Global Garbage Collection (GGC) mechanism to improve response times and reduce performance variability for a RAID array of SSDs. A coordination method is employed so that GCs in the array can run at the same time. The coordination can exhibit substantial performance improvement. In this paper, we explore various GC coordination algorithms. We develop reactive and proactive GC coordination algorithms and evaluate their I/O performance and block erase counts for various workloads. We show that a proactive GC coordination algorithm can improve the I/O response times by up to 9% further and increase the lifetime of SSDs by reducing the number of block erase counts by up to 79% compared to a reactive algorithm.

99 GENERAL AND MISCELLANEOUS↗

A Next-Generation Parallel File System Environment for the OLCF

When deployed in 2008/2009 the Spider system at the Oak Ridge National Laboratory s Leadership Computing Facility (OLCF) was the world s largest scale Lustre parallel file system. Envisioned as a shared parallel file system capable of delivering both the bandwidth and capacity requirements of the OLCF s diverse computational environment, Spider has since become a blueprint for shared Lustre environments deployed worldwide. Designed to support the parallel I/O requirements of the Jaguar XT5 system and other smallerscale platforms at the OLCF, the upgrade to the Titan XK6 heterogeneous system will begin to push the limits of Spider s original design by mid 2013. With a doubling in total system memory and a 10x increase in FLOPS, Titan will require both higher bandwidth and larger total capacity. Our goal is to provide a 4x increase in total I/O bandwidth from over 240GB=sec today to 1TB=sec and a doubling in total capacity. While aggregate bandwidth and total capacity remain important capabilities, an equally important goal in our efforts is dramatically increasing metadata performance, currently the Achilles heel of parallel file systems at leadership. We present in this paper an analysis of our current I/O workloads, our operational experiences with the Spider parallel file systems, the high-level design of our Spider upgrade, and our efforts in developing benchmarks that synthesize our performance requirements based on our workload characterization studies.

99 GENERAL AND MISCELLANEOUS↗

A Multi-Methods Approach to HRA and Human Performance Modeling: A Field Assessment

The Advanced Test Reactor (ATR) is a research reactor at the Idaho National Laboratory is primarily designed and used to test materials to be used in other, larger-scale and prototype reactors. The reactor offers various specialized systems and allows certain experiments to be run at their own temperature and pressure. The ATR Canal temporarily stores completed experiments and used fuel. It also has facilities to conduct underwater operations such as experiment examination or removal. In reviewing the ATR safety basis, a number of concerns were identified involving the ATR canal. A brief study identified ergonomic issues involving the manual handling of fuel elements in the canal that may increase the probability of human error and possible unwanted acute physical outcomes to the operator. In response to this concern, that refined the previous HRA scoping analysis by determining the probability of the inadvertent exposure of a fuel element to the air during fuel movement and inspection was conducted. The HRA analysis employed the SPAR-H method and was supplemented by information gained from a detailed analysis of the fuel inspection and transfer tasks. This latter analysis included ergonomics, work cycles, task duration, and workload imposed by tool and workplace characteristics, personal protective clothing, and operational practices that have the potential to increase physical and mental workload. Part of this analysis consisted of NASA-TLX analyses, combined with operational sequence analysis, computational human performance analysis (CHPA), and 3D graphical modeling to determine task failures and precursors to such failures that have safety implications. Experience in applying multiple analysis techniques in support of HRA methods is discussed.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

D-Factor: A Quantitative Model of Application Slow-Down in Multi-Resource Shared Systems

Scheduling multiple jobs onto a platform enhances system utilization by sharing resources. The benefits from higher resource utilization include reduced cost to construct, operate, and maintain a system, which often include energy consumption. Maximizing these benefits comes at a price - resource contention among jobs increases job completion time. In this paper, we analyze slow-downs of jobs due to contention for multiple resources in a system; referred to as dilation factor. We observe that multiple-resource contention creates non-linear dilation factors of jobs. From this observation, we establish a general quantitative model for dilation factors of jobs in multi-resource systems. A job is characterized by a vector-valued loading statistics and dilation factors of a job set are given by a quadratic function of their loading vectors. We demonstrate how to systematically characterize a job, maintain the data structure to calculate the dilation factor (loading matrix), and calculate the dilation factor of each job. We validate the accuracy of the model with multiple processes running on a native Linux server, virtualized servers, and with multiple MapReduce workloads co-scheduled in a cluster. Evaluation with measured data shows that the D-factor model has an error margin of less than 16%. We also show that the model can be integrated with an existing on-line scheduler to minimize the makespan of workloads.

99 GENERAL AND MISCELLANEOUS↗

A Case for Optimistic Coordination in HPC Storage Systems

High-performance computing (HPC) storage systems rely on access coordination to ensure that concurrent updates do not produce incoherent results. HPC storage systems typically employ pessimistic distributed locking to provide this functionality in cases where applications cannot perform their own coordination. This approach, however, introduces significant performance overhead and complicates fault handling. In this work we evaluate the viability of optimistic conditional storage operations as an alternative to distributed locking in HPC storage systems. We investigate design strategies and compare the two approaches in a prototype object storage system using a parallel read/modify/write benchmark. Our prototype illustrates that conditional operations can be easily integrated into distributed object storage systems and can outperform standard coordination primitives for simple update workloads. Our experiments show that conditional updates can achieve over two orders of magnitude higher performance than pessimistic locking for some parallel read/modify/write workloads.

Carns, Philip↗

Computational Human Performance Modeling For Alarm System Design

The introduction of new technologies like adaptive automation systems and advanced alarms processing and presentation techniques in nuclear power plants is already having an impact on the safety and effectiveness of plant operations and also the role of the control room operator. This impact is expected to escalate dramatically as more and more nuclear power utilities embark on upgrade projects in order to extend the lifetime of their plants. One of the most visible impacts in control rooms will be the need to replace aging alarm systems. Because most of these alarm systems use obsolete technologies, the methods, techniques and tools that were used to design the previous generation of alarm system designs are no longer effective and need to be updated. The same applies to the need to analyze and redefine operators’ alarm handling tasks. In the past, methods for analyzing human tasks and workload have relied on crude, paper-based methods that often lacked traceability. New approaches are needed to allow analysts to model and represent the new concepts of alarm operation and human-system interaction. State-of-the-art task simulation tools are now available that offer a cost-effective and efficient method for examining the effect of operator performance in different conditions and operational scenarios. A discrete event simulation system was used by human factors researchers at the Idaho National Laboratory to develop a generic alarm handling model to examine the effect of operator performance with simulated modern alarm system. It allowed analysts to evaluate alarm generation patterns as well as critical task times and human workload predicted by the system.

99 GENERAL AND MISCELLANEOUS↗

The Use Of Computational Human Performance Modeling As Task Analysis Tool

During a review of the Advanced Test Reactor safety basis at the Idaho National Laboratory, human factors engineers identified ergonomic and human reliability risks involving the inadvertent exposure of a fuel element to the air during manual fuel movement and inspection in the canal. There were clear indications that these risks increased the probability of human error and possible severe physical outcomes to the operator. In response to this concern, a detailed study was conducted to determine the probability of the inadvertent exposure of a fuel element. Due to practical and safety constraints, the task network analysis technique was employed to study the work procedures at the canal. Discrete-event simulation software was used to model the entire procedure as well as the salient physical attributes of the task environment, such as distances walked, the effect of dropped tools, the effect of hazardous body postures, and physical exertion due to strenuous tool handling. The model also allowed analysis of the effect of cognitive processes such as visual perception demands, auditory information and verbal communication. The model made it possible to obtain reliable predictions of operator performance and workload estimates. It was also found that operator workload as well as the probability of human error in the fuel inspection and transfer task were influenced by the concurrent nature of certain phases of the task and the associated demand on cognitive and physical resources. More importantly, it was possible to determine with reasonable accuracy the stages as well as physical locations in the fuel handling task where operators would be most at risk of losing their balance and falling into the canal. The model also provided sufficient information for a human reliability analysis that indicated that the postulated fuel exposure accident was less than credible.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Optimized Hypervisor Scheduler for Parallel Discrete Event Simulations on Virtual Machine Platforms

With the advent of virtual machine (VM)-based platforms for parallel computing, it is now possible to execute parallel discrete event simulations (PDES) over multiple virtual machines, in contrast to executing in native mode directly over hardware as is traditionally done over the past decades. While mature VM-based parallel systems now offer new, compelling benefits such as serviceability, dynamic reconfigurability and overall cost effectiveness, the runtime performance of parallel applications can be significantly affected. In particular, most VM-based platforms are optimized for general workloads, but PDES execution exhibits unique dynamics significantly different from other workloads. Here we first present results from experiments that highlight the gross deterioration of the runtime performance of VM-based PDES simulations when executed using traditional VM schedulers, quantitatively showing the bad scaling properties of the scheduler as the number of VMs is increased. The mismatch is fundamental in nature in the sense that any fairness-based VM scheduler implementation would exhibit this mismatch with PDES runs. We also present a new scheduler optimized specifically for PDES applications, and describe its design and implementation. Experimental results obtained from running PDES benchmarks (PHOLD and vehicular traffic simulations) over VMs show over an order of magnitude improvement in the run time of the PDES-optimized scheduler relative to the regular VM scheduler, with over 20 reduction in run time of simulations using up to 64 VMs. The observations and results are timely in the context of emerging systems such as cloud platforms and VM-based high performance computing installations, highlighting to the community the need for PDES-specific support, and the feasibility of significantly reducing the runtime overhead for scalable PDES on VM platforms.

Yoginath, Srikanth B↗

Reliability and performance of a system-on-a-chip by predictive wear-out based activation of functional components

A processor-implemented method for determining aging of a processing unit in a processor the method comprising: calculating an effective aging profile for the processing unit wherein the effective aging profile quantifies the effects of aging on the processing unit; combining the effective aging profile with process variation data, actual workload data and operating conditions data for the processing unit; and determining aging through an aging sensor of the processing unit using the effective aging profile, the process variation data, the actual workload data, architectural characteristics and redundancy data, and the operating conditions data for the processing unit.

97 MATHEMATICS AND COMPUTING↗

Asynchronous Object Storage with QoS for Scientific and Commercial Big Data

This paper presents our design for an asynchronous object storage system intended for use in scientific and commercial big data workloads. Use cases from the target workload do- mains are used to motivate the key abstractions used in the application programming interface (API). The architecture of the Scalable Object Store (SOS), a prototype object stor- age system that supports the API s facilities, is presented. The SOS serves as a vehicle for future research into scalable and resilient big data object storage. We briefly review our research into providing efficient storage servers capable of providing quality of service (QoS) contracts relevant for big data use cases.

Cloud Storage↗