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 541 records · Page 30

Coordinating Garbage Collection for Arrays of Solid-state Drives

Although solid-state drives (SSDs) offer significant performance improvements over hard disk drives (HDDs) for a number of workloads, they can exhibit substantial variance in request latency and throughput as a result of garbage collection (GC). When GC conflicts with an I/O stream, the stream can make no forward progress until the GC cycle completes. GC cycles are scheduled by logic internal to the SSD based on several factors such as the pattern, frequency, and volume of write requests. When SSDs are used in a RAID with currently available technology, the lack of coordination of the SSD-local GC cycles amplifies this performance variance. We propose a global garbage collection (GGC) mechanism to improve response times and reduce performance variability for a RAID of SSDs. We include a high-level design of SSD-aware RAID controller and GGC-capable SSD devices and algorithms to coordinate the GGC cycles. We develop reactive and proactive GC coordination algorithms and evaluate their I/O performance and block erase counts for various workloads. Our simulations show that GC coordination by a reactive scheme improves average response time and reduces performance variability for a wide variety of enterprise workloads. For bursty, write-dominated workloads, response time was improved by 69% and performance variability was reduced by 71%. We show that a proactive GC coordination algorithm can further improve the I/O response times by up to 9% and the performance variability by up to 15%. We also observe that it could increase the lifetimes of SSDs with some workloads (e.g. Financial) by reducing the number of block erase counts by up to 79% relative to a reactive algorithm for write-dominant enterprise workloads.

Flash Memory↗

Spatial Locality-Aware Cache Partitioning for Effective Cache Sharing, In: 2015 44th International Conference on Parallel Processing

In modern multi-core processors, last-level caches (LLCs) are typically shared among multiple cores. Previous works have shown that such sharing is beneficial as different workloads have different needs for cache capacity, and logical partitioning of capacity can improve system performance. However, what is missing in previous works on partitioning shared LLCs is that the heterogeneity in spatial locality among workloads has not been explored. In other words, all the cores use the same block/line size in shared LLCs. In this work, we highlight that exploiting spatial locality enables much more effective cache sharing. The fundamental reason is that for many memory intensive workloads, their cache capacity requirements can be drastically reduced when a large block size is employed, therefore they can effectively donate more capacity to other workloads. To leverage spatial locality for cache partitioning effectively, we first propose a simple yet effective mechanism to measure both spatial and temporal locality at run-time. The locality information is then used to determine both the proper block size and the capacity assigned to each workload. Our experiments show that our Spatial Locality-aware Cache Partitioning (SLCP) significantly outperforms the previous works. We also present several case studies that dissect the effectiveness of SLCP compared to the existing approaches.

Computer Science↗

MAPA: Multi-Accelerator Pattern Allocation Policy for Multi-Tenant GPU Servers

Multi-accelerator servers are increasingly being deployed in shared multi-tenant environments (such as in cloud data centers) in order to meet the demands of large-scale compute-intensive workloads. In addition, these accelerators are increasingly being inter-connected in complex topologies and workloads are exhibiting a wider variety of inter-accelerator communication patterns. However, existing al-location policies are ill-suited for these emerging use-cases. Specifically, this work identifies that multi-accelerator workloads are commonly fragmented leading to reduced bandwidth and increased latency for inter-accelerator communication. We propose Multi-Accelerator Pattern Allocation (MAPA), a graph pattern mining approach towards providing generalized allocation support for allocating multi-accelerator workloads on multi-accelerator servers. We demonstrate that MAPA is able to improve the execution time of multi-accelerator workloads and that MAPA is able to provide generalized benefits across various accelerator topologies. Finally, we demonstrate a speedup of 12.4% for75th percentile of jobs with the worst case execution time reduced by up to 35% against baseline policy using MAPA.

Ranganath, Kiran↗

AI4IO: A suite of AI-based tools for IO-aware scheduling

Traditional workload managers do not have the capacity to consider how IO contention can increase job runtime and even cause entire resource allocations to be wasted. Whether from bursts of IO demand or parallel file systems (PFS) performance degradation, IO contention must be identified and addressed to ensure maximum performance. In this paper, we present AI4IO (AI for IO), a suite of tools using AI methods to prevent and mitigate performance losses due to IO contention. AI4IO enables existing workload managers to become IO-aware. Currently, AI4IO consists of two tools: PRIONN and CanarIO. PRIONN predicts IO contention and empowers schedulers to prevent it. CanarIO mitigates the impact of IO contention when it does occur. We measure the effectiveness of AI4IO when integrated into Flux, a next-generation scheduler, for both small- and large-scale IO-intensive job workloads. Our results show that integrating AI4IO into Flux improves the workload makespan up to 6.4%, which can account for more than 18,000 node-h of saved resources per week on a production cluster in our large-scale workload.

Wyatt, II, Michael R.↗

Crew procedures for microwave landing system operations

The objective of this study was to identify crew procedures involved in Microwave Landing System (MLS) operations and to obtain a preliminary assessment of crew workload. The crew procedures were identified for three different complements of airborne equipment coupled to an autopilot. Using these three equipment complements, crew tasks were identified for MLS approaches and precision departures and compared to an ILS approach and a normal departure. Workload comparisons between the approaches and departures were made by using a task-timeline analysis program that obtained workload indexes, i.e., the radio of time required to complete the tasks to the time available. The results showed an increase in workload for the MLS scenario for one of the equipment complements. However, even this workload was within the capacity of two crew members.

Summers, Leland G.↗

Performance analysis of the Alliant FX/8 multiprocessor using statistical clustering

Results for two distinct, real, scientific workloads executed on an Alliant FX/8 are discussed. A combination of user concurrency and system overhead measurements was taken for both workloads. Preliminary analysis shows that the first sampled workload is comprised of consistently high user concurrency, low system overhead, and little paging. The second sample has much less user concurrency, but significant paging and system overhead. Statistical cluster analysis is used to extract a state transition model to jointly characterize user concurrency and system overhead. A skewness factor is introduced and used to bring out the effects of unbalanced clustering when determining states with important transitions. The results from the models show that during the collection of the first sample, the system was operating in states of high user concurrency approximately 75 percent of the time. The second workload sample shows the system in high user concurrency states only 26 percent of the time. In addition, it is ascertained that high system overhead is usually accompanied by low user concurrency. The analysis also shows a high predictability of system behavior for both workloads.

Dimpsey, Robert Tod↗

An analysis of scatter decomposition

A formal analysis of a mapping method known as scatter decomposition (SD) is presented. SD divides an irregular domain into many equal-size pieces and distributes them modularly among processors. It is shown that, if a correlation in workload is a convex function of distance, then scattering a more finely decomposed domain yields a lower average processor workload variance; if the workload process is stationary Gaussian and the correlation function decreases linearly in distance to zero and then remains zero, scattering a more finely decomposed domain yields a lower expected maximum processor workload. Finally, if the correlation function decreases linearly across the entire domain, then (among all mappings that assign an equal number of domain pieces to each processor) SD minimizes the average processor workload variance. The dependence of these results on the assumption of decreasing correlation is illustrated with cases where a coarser granularity actually achieves better load balance.

Nicol, David M.↗

A Comparison of a Brain-Based Adaptive System and a Manual Adaptable System for Invoking Automation

Two experiments are presented that examine alternative methods for invoking automation. In each experiment, participants were asked to perform simultaneously a monitoring task and a resource management task as well as a tracking task that changed between automatic and manual modes. The monitoring task required participants to detect failures of an automated system to correct aberrant conditions under either high or low system reliability. Performance on each task was assessed as well as situation awareness and subjective workload. In the first experiment, half of the participants worked with a brain-based system that used their EEG signals to switch the tracking task between automatic and manual modes. The remaining participants were yoked to participants from the adaptive condition and received the same schedule of mode switches, but their EEG had no effect on the automation. Within each group, half of the participants were assigned to either the low or high reliability monitoring task. In addition, within each combination of automation invocation and system reliability, participants were separated into high and low complacency potential groups. The results revealed no significant effects of automation invocation on the performance measures; however, the high complacency individuals demonstrated better situation awareness when working with the adaptive automation system. The second experiment was the same as the first with one important exception. Automation was invoked manually. Thus, half of the participants pressed a button to invoke automation for 10 s. The remaining participants were yoked to participants from the adaptable condition and received the same schedule of mode switches, but they had no control over the automation. The results showed that participants who could invoke automation performed more poorly on the resource management task and reported higher levels of subjective workload. Further, those who invoked automation more frequently performed more poorly on the tracking task and reported higher levels of subjective workload. and the adaptable condition in the second experiment revealed only one significant difference: the subjective workload was higher in the adaptable condition. Overall, the results show that a brain-based, adaptive automation system may facilitate situation awareness for those individuals who are more complacent toward automation. By contrast, requiring operators to invoke automation manually may have some detrimental impact on performance but does appear to increases subjective workload relative to an adaptive system.

Bailey, Nathan R.↗

Pilot Subjective Assessments During an Investigation of Separation Function Allocation Using a Human-In-The-Loop Simulation

Two human-in-the-loop simulation experiments were conducted to investigate allocation of separation assurance functions between ground and air and between humans and automation. The experiments modeled a mixed-operations concept in which aircraft receiving ground-based separation services shared the airspace with aircraft providing their own separation service (i.e., self-separation). The two experiments, one pilot-focused and the other controller-focused, addressed selected key issues of mixed operations and modeling an emergence of NextGen technologies and procedures. This paper focuses on the results of the subjective assessments of pilots collected during the pilot-focused human-in-the-loop simulation, specifically workload and situation awareness. Generally the results revealed that across all conditions, pilots' perceived workload was low to medium, with the highest reported levels of workload occurring when the pilots experienced a loss of separation during the scenario. Furthermore, the results from the workload data and situation awareness data were complimentary such that when pilots reported lower levels of workload they also experienced higher levels of situation awareness.

Burke, Kelly A.↗

How do Air Traffic Controllers Use Automation and Tools Differently During High Demand Situations?

In a human-in-the-loop simulation, two air traffic controllers managed identical airspace while burdened with higher than average workload, and while using advanced tools and automation designed to assist with scheduling aircraft on multiple arrival flows to a single meter fix. This paper compares the strategies employed by each controller, and investigates how the controllers' strategies change while managing their airspace under more normal workload conditions and a higher workload condition. Each controller engaged in different methods of maneuvering aircraft to arrive on schedule, and adapted their strategies to cope with the increased workload in different ways. Based on the conclusions three suggestions are made: that quickly providing air traffic controllers with recommendations and information to assist with maneuvering and scheduling aircraft when burdened with increased workload will improve the air traffic controller's effectiveness, that the tools should adapt to the strategy currently employed by a controller, and that training should emphasize which traffic management strategies are most effective given specific airspace demands.

ATC↗

Monitoring Airspace Complexity and Determining Contributing Factors

The national airspace has evolved over many years to accommodate increased traffic demand [1] while simultaneously maintaining one of the safest forms of transportation [2], [3]. One of the reasons for this success is the ability of the system and the operators to adapt and accommodate to situations that routinely disrupt optimal operations. These situations may include: adverse weather, delays, early arrivals, equipment outages, and other factors that are outside the operators ability to control. These factors can lead to states where automation is unable to properly handle these issues and therefore air traffic controllers and pilots have to intervene, ultimately increasing communication between operators resulting in higher workload. As controller workload increases to handle sub-optimal operating conditions this can be viewed as an increase in complexity. The reasoning for this is because humans are now required to make tactical decisions in response to external factors, resulting in a departure from the strategic plan where operations would be more efficiently managed. Human operators control airspace complexity under rigid regulations that are constantly changing. The airspace is divided into sectors and the number of aircraft assigned to each controller is limited for safe handling. There has been past work that devised airspace complexity metrics in commercial aviation and related these metrics to controller workload (e.g., [4],[5]). The upper bounds on the system load are pre-determined. Such bounds on complexity make for a safe system, but the system cannot scale and adapt to autonomous, dense, and heterogeneous traffic, including the many types of Unmanned Aerial Vehicles (UAVs) envisioned to be added to the operations. We hypothesize that, as traffic density and heterogeneity grow, and other key metrics change, there will be phase transitions at which the way traffic should be managed changes significantly [6]. We offer a method for in-time detection of contributing factors that lead to phase transitions, characterized by increased complexity. To the best of our knowledge, there is no tool similar to our proposed effort that identifies such contributing factors or precursor patterns. To define the scope we are proposing to measure complexity from the viewpoint of the Terminal Radar Approach Control Facilities (TRACON) controller’s perspective. In particular we are analyzing arrivals into KSFO. With safety as the top concern for airspace operators, it is important to recognize that as density and heterogeneity grow, the focus of the system will change. Times of the day when the airspace has low density and heterogeneity, the flights will follow more efficient paths where the aircraft move on established routes that are more or less directly to the destination. However, when density and heterogeneity increases, the system will begin changing focus to avoiding conflicts and collisions and route the flights in a more flexible way. Higher flexibility requires more communication and coordination between controllers and pilots which the current automation is unable to handle. This paper proposes a novel approach that monitors airspace complexity at multiple scales, uses a Machine Learning-based tool that predicts when operations will transition to a regime of greater complexity, and identifies actions that can reduce the complexity while still maintaining efficient and safe operations. We demonstrate our proposed approach using data from multiple complementary sources. This includes, but is not limited to: historical aircraft surveillance data from NASA’s Sherlock Data Warehouse [7], METAR weather data, and airport configuration data from Aviation System Performance Metrics (ASPM). The surveillance data flight paths are sampled at a variable sample rate — increasing as the aircraft approaches the airport. This is due to how Sherlock manages flight track stitching between different radar facilities which have different sampling rates. The weather and performance data are logged at defined intervals throughout the day at a courser refresh rate. In addition to the logged data and metrics, we leverage pre-defined Standard Terminal Arrival Routes (STARs) procedures to characterize the path of each flight. Each flight files for one of these routes in the flight plan well before entering the terminal airspace, and approximately follows the route until it leaves the STAR, typically on the final fix of a runway transition. However, most flights do not always fly the full STAR procedure to completion [8], but the majority do adhere to the fixes within the common route of the procedure. Our approach leverages fixes in the common route of each of the STARs to build a reference path to the airport. This allows us to characterize the flight paths in what we are defining as the “maneuvering area” (the airspace between the STAR and before the flight is lined up on the runway’s final approach) to determine how off nominal the flights are to calculate its complexity score. Determining airspace complexity is a concept that does not have a concrete answer. In designing this metric, we consider what increases the workload for the air traffic controllers. Consequently more specialized vectoring maneuvers results in higher workload. Accordingly, we start with a theory: each flight has a direct path it takes from the STAR’s common route to the final approach’s outer marker fix for the flight’s landing runway. It is important to note that the direct path is only used as a reference. If the majority of the flights have a large consistent offset as compared to other routes it does not necessarily mean that those flights have higher complexity. We are merely building a distribution based on this direct path for that particular STAR and runway pair to determine the normal mode of operations for that route. Flights that are in the upper tail of these distributions will result in higher complexity scores and flights that fly in the median will represent the normal mode of operations and therefore will have lower complexity scores. Since flights following each STAR route take different paths to the airport, we have a different distribution for each STAR route and therefore can model these distributions to compute a complexity score from their respective normalized distributions. To evaluate the effectiveness of our proposed airspace complexity metric we will compare against an established approach based on trajectory clustering [9]. This unsupervised learning technique consists of the following steps: (1) identify the general maneuvering areas (waypoints) by performing $\kappa$-means or DBSCAN clustering on locations where aircraft frequently turn based on the surveillance radar track data, (2) map flight trajectories onto sequences of waypoints, and (3) cluster the sequences based on their common subsequences. From a high-level perspective, this baseline model learns nominal operations in the airspace through the sequence of waypoints that are representative of where aircraft change direction and defines deviations from the nominal operations as “complex.” Therefore, more deviations from the nominal operations correspond to higher complexity values. For our validation, we re-implemented this technique and tune model hyper-parameters to correctly detect waypoints for the arrival traffic into the San Francisco bay area. We will compute the complexity measure over a one-year period using our proposed technique as well as the baseline. Our validation will be based on each technique’s ability to detect a set of undesirable outcomes (e.g., go-arounds, holding patterns, average time in the airspace, etc.). Since our current complexity metric is derived from the offset from the direct reference path, it’s important to understand what causes these offsets. In many of the flights with high offset distance, flights performing holding patterns and S turns can be observed. These maneuvering tactics are utilized to add distance between the aircraft and the destination runway to prevent multiple flights from having conflicting arrival times. In order to predict a rise in complexity (or the precursor to complexity), it’s necessary to be able to identify these potential conflicts (which in turn, result in higher offsets). To do this, we define a “representative flight” for each STAR route and runway pair. This flight is approximately the path the flight would take if there was a clear path with no other flights in the airspace — including the time remaining to the airport. We first identify the flights for a given STAR runway pair using the offset to the reference path distributions that fall between the 44-55 percentiles. This yields the flights that conform to the most normal mode of operation. Each of these flights is partitioned based on the percent complete from the entry point into the maneuvering areas from 0\% – 100\% complete. Then for each percent “bin”, we take the median value of the flight’s latitude/longitude coordinates, airspeed, and (non causal) time remaining to the airport to construct a lookup table for each percent complete bin on a given route. As a flight enters the maneuvering area, we can find the estimated arrival time of a flight to the airport by finding the closest point to the representative path’s percent complete bin (relative to the flight’s current position at any snapshot in the airspace) and therefore retrieve the corresponding remaining time left on the “representative path”. We assume that the flight will follow the representative path to completion when deriving these estimates. We can then compare these estimated arrival times against other flights for the same snapshot in time to identify potential conflicts. If more flights are estimated to arrive within a tolerance window than there are runways available, then we have a potential conflict. We can use this derived measure along with other factors expected to add disruption to the operation such as weather and runway configuration changes as an input to machine learning tools to detect precursors that increases in our complexity measure. This novel method will assist in uncovering insights into the contributing factors that lead to increased complexity that may allow for in-time responses to avoid reaching a high complexity state in the airspace.

complexity↗

EXPLORING THE POTENTIAL OF SHORT-TIME FOURIER TRANSFORMS FOR ANALYZING SKIN CONDUCTANCE AND PUPILLOMETRY IN REAL-TIME APPLICATIONS

The development of real-time predictors of mental workload is critical for the practical application of augmented cognition to human-machine systems. This paper explores a novel method based on a short-time Fourier transform (STFT) for analyzing galvanic skin conductance (SC) and pupillometry time-series data to extract estimates of mental workload with temporal bandwidth high-enough to be useful for augmented cognition applications. We tested the method in the context of a process control task based on the DURESS simulation developed by Vincente and Pawlak (1994; ported to Java by Cosentino,& Ross, 1999). SC, pupil dilation, blink rate, and visual scanning patterns were measured for four participants actively engaged in controlling the simulation. Fault events were introduced that required participants to diagnose errors and make control adjustments to keep the simulator operating within a target range. We were interested in whether the STFT of these measures would produce visible effects of the increase in mental workload and stress associated with these events. Graphical exploratory data analysis of the STFT showed visible increases in the power spectrum across a range of frequencies directly following fault events. We believe this approach shows potential as a relatively unobtrusive, low-cost, high bandwidth measure of mental workload that could be particularly useful for the application of augmented cognition to human-machine systems.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Human System Simulation in Support of Human Performance Technical Basis at NPPs

This paper focuses on strategies and progress toward establishing the Idaho National Laboratory’s (INL’s) Human Systems Simulator Laboratory at the Center for Advanced Energy Studies (CAES), a consortium of Idaho State Universities. The INL is one of the National Laboratories of the US Department of Energy. One of the first planned applications for the Human Systems Simulator Laboratory is implementation of a dynamic nuclear power plant simulation (NPP) where studies of operator workload, situation awareness, performance and preference will be carried out in simulated control rooms including nuclear power plant control rooms. Simulation offers a means by which to review operational concepts, improve design practices and provide a technical basis for licensing decisions. In preparation for the next generation power plant and current government and industry efforts in support of light water reactor sustainability, human operators will be attached to a suite of physiological measurement instruments and, in combination with traditional Human Factors Measurement techniques, carry out control room tasks in simulated advanced digital and hybrid analog/digital control rooms. The current focus of the Human Systems Simulator Laboratory is building core competence in quantitative and qualitative measurements of situation awareness and workload. Of particular interest is whether introduction of digital systems including automated procedures has the potential to reduce workload and enhance safety while improving situation awareness or whether workload is merely shifted and situation awareness is modified in yet to be determined ways. Data analysis is carried out by engineers and scientists and includes measures of the physical and neurological correlates of human performance. The current approach supports a user-centered design philosophy (see ISO 13407 “Human Centered Design Process for Interactive Systems, 1999) wherein the context for task performance along with the requirements of the end-user are taken into account during the design process and the validity of design is determined through testing of real end users

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Evaluating the potential of disaggregated memory systems for HPC applications

Summary Disaggregated memory is a promising approach that addresses the limitations of traditional memory architectures by enabling memory to be decoupled from compute nodes and shared across a data center. Cloud platforms have deployed such systems to improve overall system memory utilization, but performance can vary across workloads. High‐performance computing (HPC) is crucial in scientific and engineering applications, where HPC machines also face the issue of underutilized memory. As a result, improving system memory utilization while understanding workload performance is essential for HPC operators. Therefore, learning the potential of a disaggregated memory system before deployment is a critical step. This paper proposes a methodology for exploring the design space of a disaggregated memory system. It incorporates key metrics that affect performance on disaggregated memory systems: memory capacity, local and remote memory access ratio, injection bandwidth, and bisection bandwidth, providing an intuitive approach to guide machine configurations based on technology trends and workload characteristics. We apply our methodology to analyze thirteen diverse workloads, including AI training, data analysis, genomics, protein, fusion, atomic nuclei, and traditional HPC bookends. Our methodology demonstrates the ability to comprehend the potential and pitfalls of a disaggregated memory system and provides motivation for machine configurations. Our results show that eleven of our thirteen applications can leverage injection bandwidth disaggregated memory without affecting performance, while one pays a rack bisection bandwidth penalty and two pay the system‐wide bisection bandwidth penalty. In addition, we also show that intra‐rack memory disaggregation would meet the application's memory requirement and provide enough remote memory bandwidth.

Ding, Nan↗

An automated and portable method for selecting an optimal GPU frequency

Power consumption poses a significant challenge in current and emerging graphics processing unit (GPU) enabled high-performance computing systems. In modern GPUs, dynamic voltage frequency scaling (DVFS) appears to be a reliable control to regulate power consumption and performance. However, the DVFS design space is large - hence, brute-force approaches are infeasible to select the optimal frequency. Furthermore, no single frequency can be universally optimal for applications with varying computational intensities. Thus, the application's complexity and the availability of a wide range of frequency settings are a challenge in selecting the optimal frequency configuration for a given GPU workload. To that end, this paper proposes a systematic approach that consists of three steps. The feature characterization study identifies the fine-grain GPU utilization metrics that influence the power consumption and execution time of a given workload. To understand the performance, power, and energy consumption behaviors of a workload across GPU's DVFS design space, we derived analytical power and performance models using the identified fine-grain features. Here, it is shown that the same set of GPU utilization metrics can estimate both the power consumption and execution time while being agnostic of changes to frequency and input sizes. Applying a power control with the single objective of reducing power may cause performance degradation, leading to more energy consumption. A multi-objective approach is proposed to select the optimal GPU DVFS configuration for a workload that reduces power consumption with negligible degradation in performance. The evaluation was conducted using SPEC ACCEL benchmarks and three real applications - NAMD LAMMPS, and LSTM on NVIDIA GV100, GA100, and AMD MI210 GPUs. On average, real applications showed 29.6% energy savings with a performance loss of 5.2% on GA100 and 22.6% energy savings with a performance loss of 4.7% on GV100. Moreover, the proposed models are portable to real applications, GPU architectures, and vendors, and require metric collection at only the default frequency rather than all supported DVFS configurations. Additionally, we conducted a comparison between our models and the GPU assembly instructions (PTX)-based static models. The results revealed a significant reduction in the average error rates, with a decrease from 19.7% to 3.1% for power models and from 29.4% to 5.2% for performance models.

97 MATHEMATICS AND COMPUTING↗

Evaluating adaptive and predictive power management strategies for optimizing visualization performance on supercomputers

Power is becoming an increasingly scarce resource on the next generation of supercomputers, and should be used wisely to improve overall performance. One strategy for improving power usage is hardware overprovisioning, i.e., systems with more nodes than can be run at full power simultaneously without exceeding the system-wide power limit. With this study, we compare two strategies for allocating power throughout an overprovisioned system – adaptation and prediction – in the context of visualization workloads. While adaptation has been suitable for workloads with more regular execution behaviors, it may not be as suitable on visualization workloads, since they can have variable execution behaviors. This study considers a total of 104 experiments, which vary the rendering workload, power budget, allocation strategy, and node concurrency, including tests processing data sets up to 1 billion cells and using up to 18,432 cores across 512 nodes. Overall, we find that prediction is a superior strategy for this use case, improving performance up to 27% compared to an adaptive strategy.

97 MATHEMATICS AND COMPUTING↗

Preparation of the Multi-Site Data Processing at the Vera C. Rubin Observatory

The Vera C. Rubin Observatory’s Legacy Survey of Space and Time (LSST) Camera is scheduled to start taking data in the summer of 2025. The Data Release Production will run the LSST Science Pipe software at data facilities in the US, France and the UK. The LSST Science Pipeline consists of complex directed acyclic graphs (DAGs) of tasks. Rubin will use the Production and Distributed Analysis (PanDA) workflow and workload management system to orchestrate this complex workflow and the distribution of workloads to the data facilities. When run end-to-end by a team of data production staff, this processing (the Science Pipelines, distributed by the workflow and workload management system) is referred to as a 'campaign'. This paper describes the central services and data facility specific services that support this multi-site data process model, including the service deployment infrastructure, the workload and workflow system, the Campaign Management tools, and connection to Rubin Data Management. This paper will also mention the experience of processing the Rubin Commissioning Camera data. All these are part of the effort to scale up the processing capabilities for the expected very large data volume from the LSST Camera.

Yang, Wei [SLAC]↗

Addressing Load Imbalance in Bioinformatics and Biomedical Applications: Efficient Scheduling across Multiple GPUs

Computational bioinformatics and biomedical applications frequently contain heterogeneously sized units of work or tasks, for instance due to variability in the sizes of biological sequences and molecules. Variable-sized workloads lead to load imbalances in parallel implementations which detract from efficiency and performance. Many modern computing resources now have multiple graphics processing units(GPUs) per computer for acceleration. These multiple GPU resources need to be used efficiently through balancing of workloads across the GPUs. OpenMP is a portable directive-based parallel programming API used ubiquitously in bioscience applications to program CPUs; recently, the use of OpenMP directives for GPU acceleration has become possible. Here, motivated by experiences with imbalanced loads in GPU-accelerated bioinformatics applications, we address the load balancing problem using OpenMP task-to-GPU scheduling combined with OpenMP GPU offloading for multiply heterogeneous workloads – loads with both variable input sizes, and simultaneously, variable convergence rates for algorithms with a stochastic component – scheduled across multiple GPUs. We aim to develop strategies which are both easy to use and have lower overheads, and may be incorporated incrementally in existing programs which already make use of OpenMP for CPU-based threading in order to make use of multi-GPU computers. We test different combinations of input size variability and convergence rate variability, and characterize the effects of these different scenarios on the performance of scheduling strategies across multiple GPUs with OpenMP. We present several dynamic scheduling solutions for different parallel patterns, explore optimizations, and provide publicly available example computational kernels to make these strategies easy to use in programs. This work will enable application developers to efficiently and easily use multiple GPUs for imbalanced workloads found in bioinformatics and biomedical applications.

Thavappiragasam, Mathialakan↗