Search NASA⌕ Search

SEARCH · Search NASA

Results for “runtime performance”

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
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Development and Validation of Algorithms That Analyze Communicating Thermostat Data to Identify Enclosure Retrofit Opportunities

Annual energy savings of up to $\$ 4$ to $\$ 5$ billion could be achieved nationwide through basic insulation and heating system retrofits of existing homes. However, current utility energy efficiency programs are costly and challenging to scale. Customer acquisition occurs primarily through energy bill mailers, mass media, and online advertising that lack specificity about home-specific retrofit opportunities, expected energy savings, and cost-effectiveness. Specific retrofit opportunities are identified via on-site home energy assessments (HEAs) that are inconvenient to homeowners, expensive, and of variable accuracy. We developed computational algorithms that automatically analyze communicating thermostat (CT) heating data that could be used to increase the customer uptake of insulation and air sealing energy conservation measures (ECMs) by identifying homes with the most significant retrofit opportunities, estimating post-retrofit energy savings, and formulating home-specific outreach. The algorithms are based on an extended second-order grey-box model that characterizes a building’s thermal response using lumped elements, coupled with an empirical model of infiltration that accounts for both wind and stack effects. The basic parameters of the model correspond to actual physical parameters of the home, i.e., the home’s overall R-value of and the building envelope ACH50. Unlike the conventional approach, which estimates model parameters based on the best fit to the observed time-dependent room temperature, our approach derives correlations between the daily heating system runtime and temperature difference (indoor-outdoor) that are more robust to data quality issues in real-world applications. We also used HEA data for algorithm development and validation. With the help of our utility partners, Eversource and National Grid, we obtained data sets for hundreds of Massachusetts homes. For each home, these data sets included three sets of information anonymized by the utility: (1) CT data (HVAC runtime, room temperature, and, for some vendors, outdoor temperature and wind speed) collected by the CT vendor (one of three) over a heating season, (2) HEA report performed by the HEA vendor (same vendor for all homes), (3) Monthly utility gas bills coincident with the CT data (3 to 24 per home, depending on availability). For some homes, we also obtained blower-door test results. Initially, we applied the algorithms developed to homes with a single CT and then extended them to homes with two CTs by using an equivalent home approach. Finally, we developed algorithms for prediction of energy savings and a methodology of comparing our predictions with those generated by HEAs. The main technical results indicate that we can reliably identify homes with insulation and/or air sealing retrofit opportunities and provide accurate savings predictions. Our hypothesis is that the algorithms could be applied to utility energy efficiency programs to identify homes that could realize significant energy savings from insulation and/or air sealing retrofits. This information could then be used to reach out to those homes with highly customized outreach, thereby delivering increased program energy savings and cost-effectiveness. This would: Significantly increase the uptake rate of on-site HEAs, and Significantly increase the fraction of HEAs resulting in ECM implementation. To test these hypotheses, we designed and conducted a randomized controlled trial (RCT). The RCT results suggest that personal messaging leads to a two- to five-fold increase in the HEA uptake rate.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2025

Experience with the alpaka performance portability library in the CMS software

ion Library for Parallel Kernel Acceleration) is a header-only C++ library that provides performance portability across different back-ends, abstracting the underlying levels of parallelism. It supports serial and parallel execution on CPUs, and extremely parallel execution on NVIDIA, AMD and Intel GPUs.This contribution will show how alpaka is used in the CMS software to develop and maintain a single code base; to use different toolchains to build the code for each supported back-end, and link them into a single application; to seamlessly select the best backend at runtime, and implement portable reconstruction algorithms that run efficiently on CPUs and GPUs from different vendors. It will describe the validation and deployment of the alpaka-based implementation in the CMS High Level Trigger, and highlight how it achieves near-native performance.

Alawieh, Jaafar [CERN]↗
DOE OSTI2020

Locality-Aware Scheduling for Scalable Heterogeneous Environments

Heterogeneous computing promise boost performance of scientific applications by allowing massively parallel execution of computational tasks. However, manually managing extremely heterogeneous, multi-device systems is complicated and may result in sub-optimal performance. Specifically, data management is an extremely challenging problem on multi-device systems. In this work, we introduce two locality-aware schedulers for the Minos Computing Library (MCL), an asynchronous, task-based programming model and runtime for extremely heterogeneous systems. The first scheduler implements a pure locality-aware algorithm to maximize data reuse, though it might incur in ”hot-spots” that limit system utilization. The second scheduler mitigates this drawback by dynamically targeting between locality-awareness and system utilization based on the current workload and available computing devices. Our results show that locality-awareness greatly benefit applications that exhibit data reuse, providing up to 6.9x and 7.9x over the original MCL scheduler and equivalent OpenCL implementations, respectively. Moreover, our schedulers introduce negligible overhead compared with the original MCL scheduler and achieve similar performance for applications that don’t benefit from data locality.

Architecture, co-design, Task-based programming mo↗
DOE OSTI2020

Pulsed Thermal Tomography Nondestructive Examination of Additively Manufactured Reactor Materials (Second Annual Progress Report)

Additive manufacturing (AM) is an emerging method for cost-efficient fabrication of nuclear reactor parts. AM of metallic structures for nuclear energy applications is currently based on laser powder bed fusion (LPBF) process, which can introduce internal material flaws, such as pores and anisotropy. Integrity of AM structures needs to be evaluated nondestructively because material flaws could lead to premature failures due to exposure to high temperature, radiation and corrosive environment in a nuclear reactor. Thermal tomography (TT) provides a capability for non-destructive evaluation of sub-surface defects in arbitrary size structures. Thermal tomography is a computational method for heat diffusion-based imaging of solids, which provides 3D visualization of data from flash thermography measurements. We investigate thermal tomography imaging and nondestructive evaluation of stainless steel and nickel super alloy metallic structures produced with laser powder bed fusion (LPBF) additive manufacturing (AM) process. Metallic structures produced with LPBF contain defects, and there are limited capabilities to evaluate these structures non-destructively. Thermal tomography reconstruction of 3D apparent spatial effusivity provides information about AM structure geometry and internal material flaws. We study performance of thermal tomography in imaging of metallic structures through COMSOL computer simulations of transient heat transfer, and through reconstruction of data obtained from experimental measurements. Reconstruction of internal defects is investigated using a stainless steel 316L specimen with flat bottom hole (FBH) indentations, and Inconel 718 plate produced with laser powder bed fusion (LPBF) method, which contains imprinted hemispherical shape low density regions containing non-sintered metallic powder. The FBH’s have the same sizes as the imprinted defects in the LPBF specimens, but offer better imaging contrast. Thermal tomography reconstructions provide visualizations of internal defects, and allow for estimation of their sizes and locations. Detection sensitivity of TT is limited by noises. We investigate separation of signal from noise in thermography images using several machine learning (ML) methods, including new spatio-temporal blind source separation (STBSS) and spatio-temporal sparse dictionary learning (STSDL) methods. Performance of the ML methods is benchmarked using thermography data obtained from imaging stainless steel 316L and Inconel 718 specimens produced LPBF method with imprinted calibrated porosity defects. The ML methods are ranked by F-score and execution runtime. Finally, we investigate TT of AM stainless steel 316L specimen with imprinted internal porosity defects using relatively low-cost, small form factor infrared (IR) camera based on uncooled micro bolometer detector. Sparse coding related K-means singular value decomposition (SVD) machine learning, image processing algorithms are developed to improve quality of TT images through removal of Additive white Gaussian noise without blurring the images. Following initial qualification of an AM component for deployment in a nuclear reactor, a compact TT can also be used for in-service nondestructive evaluation (NDE). With capability to perform in-service NDE of the AM component lifecycle, TT data can be used for development of a component digital twin.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2026

Intercomparison of flood inundation models across land use types and hydrological flood stages

Flood Inundation Mapping (FIM) model selection is a key operational decision because accurate, rapid mapping underpins early warning and resource allocation. FIM performance is context-dependent and can vary with hydrograph phase, land-use/land-cover (LULC), and the evaluation benchmark. Intercomparison studies typically assess a single near-peak snapshot against one reference dataset. Here, we provide a context-stratified intercomparison across (i) multiple hydrograph phases, (ii) LULC classes, and (iii) benchmark types, for five FIM approaches spanning a wide range of physical complexity and operational cost (TRITON, LISFLOOD-FP, HEC-RAS 2D, ARC-Curve2Flood, and OWP HAND-FIM). We use the Hurricane Matthew flood (2016) in the Neuse River Basin, North Carolina, USA, as a case study. Using high-resolution remote sensing-derived flood inundation maps, hand-labeled points, and building footprints, we assess model skill across two rising and two falling hydrograph limbs and across major LULC types. Results show that model rankings shift systematically across contexts: LISFLOOD-FP ranks highest in three of four flood phases, while TRITON leads during one rising limb phase; LISFLOOD-FP performs best in vegetated areas, whereas HEC-RAS improves relative performance in agricultural and urban areas; and benchmark choice influences conclusions, with LISFLOOD-FP performing best for flooded-building detection in the late falling limb, while TRITON ranks highest against hand-labeled points. We also report representative wall-clock runtimes for each workflow to provide use-case context for operational feasibility. Together, these results offer transferable guidance for model selection and for designing large-scale, benchmark-aware FIM intercomparison studies.

Nikrou, Parvaneh [University of Alabama]↗
DOE OSTI2017

Aces4: A Platform for Computational Chemistry Calculations with Extremely Large Block-Sparse Arrays

Aces4 is a parallel programming platform comprising a DSL for Computational Chemistry and its runtime system. It offers a convenient way to express parallelism together with extensive support for extremely large, possibly sparse, distributed arrays. It aids scientists in the creation of performant, scalable, massively parallel programs that can effectively take advantage of leadership class computing systems to address important scientific questions. Aces4 has enabled the development and implementation of new methods in electronic structure theory which are breaking new ground in their ability to perform highly accurate calculations on ever larger molecular systems. In this paper the design of Aces4, which is based on the the Super Instruction Architecture approach, is described. Experimental scaling results for Molecular Cluster Perturbation Theory, a new method enabled by Aces4, and CCSD, a widely used computational chemistry method are given.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

StructuredFuzzer: Fuzzing Structured Text-Based Control Logic Applications

Rigorous testing methods are essential for ensuring the security and reliability of industrial controller software. Fuzzing, a technique that automatically discovers software bugs, has also proven effective in finding software vulnerabilities. Unsurprisingly, fuzzing has been applied to a wide range of platforms, including programmable logic controllers (PLCs). However, current approaches, such as coverage-guided evolutionary fuzzing implemented in the popular fuzzer American Fuzzy Lop Plus Plus (AFL++), are often inadequate for finding logical errors and bugs in PLC control logic applications. They primarily target generic programming languages like C/C++, Java, and Python, and do not consider the unique characteristics and behaviors of PLCs, which are often programmed using specialized programming languages like Structured Text (ST). Furthermore, these fuzzers are ill suited to deal with complex input structures encapsulated in ST, as they are not specifically designed to generate appropriate input sequences. This renders the application of traditional fuzzing techniques less efficient on these platforms. To address this issue, this paper presents a fuzzing framework designed explicitly for PLC software to discover logic bugs in applications written in ST specified by the IEC 61131-3 standard. The proposed framework incorporates a custom-tailored PLC runtime and a fuzzer designed for the purpose. We demonstrate its effectiveness by fuzzing a collection of ST programs that were crafted for evaluation purposes. We compare the performance against a popular fuzzer, namely, AFL++. The proposed fuzzing framework demonstrated its capabilities in our experiments, successfully detecting logic bugs in the tested PLC control logic applications written in ST. On average, it was at least 83 times faster than AFL++, and in certain cases, for example, it was more than 23,000 times faster.

47 OTHER INSTRUMENTATION↗
DOE OSTI2015

Proof-of-Concept Demonstrations for Computation-Based Human Reliability Analysis. Modeling Operator Performance During Flooding Scenarios

The United States (U.S.) Department of Energy (DOE) Light Water Reactor Sustainability (LWRS) program has the overall objective to help sustain the existing commercial nuclear power plants (NPPs). To accomplish this program objective, there are multiple LWRS “pathways,” or research and development (R&D) focus areas. One LWRS focus area is called the Risk-Informed Safety Margin and Characterization (RISMC) pathway. Initial efforts under this pathway to combine probabilistic and plant multi-physics models to quantify safety margins and support business decisions also included HRA, but in a somewhat simplified manner. HRA experts at Idaho National Laboratory (INL) have been collaborating with other experts to develop a computational HRA approach, called the Human Unimodel for Nuclear Technology to Enhance Reliability (HUNTER), for inclusion into the RISMC framework. The basic premise of this research is to leverage applicable computational techniques, namely simulation and modeling, to develop and then, using RAVEN as a controller, seamlessly integrate virtual operator models (HUNTER) with 1) the dynamic computational MOOSE runtime environment that includes a full-scope plant model, and 2) the RISMC framework PRA models already in use. The HUNTER computational HRA approach is a hybrid approach that leverages past work from cognitive psychology, human performance modeling, and HRA, but it is also a significant departure from existing static and even dynamic HRA methods. This report is divided into five chapters that cover the development of an external flooding event test case and associated statistical modeling considerations.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2015

Exploiting communication concurrency on high performance computing systems

Although logically available, applications may not exploit enough instantaneous communication concurrency to maximize hardware utilization on HPC systems. This is exacerbated in hybrid programming models such as SPMD+OpenMP. We present the design of a "multi-threaded" runtime able to transparently increase the instantaneous network concurrency and to provide near saturation bandwidth, independent of the application configuration and dynamic behavior. The runtime forwards communication requests from application level tasks to multiple communication servers. Our techniques alleviate the need for spatial and temporal application level message concurrency optimizations. Experimental results show improved message throughput and bandwidth by as much as 150% for 4KB bytes messages on InfiniBand and by as much as 120% for 4KB byte messages on Cray Aries. For more complex operations such as all-to-all collectives, we observe as much as 30% speedup. This translates into 23% speedup on 12,288 cores for a NAS FT implemented using FFTW. We also observe as much as 76% speedup on 1,500 cores for an already optimized UPC+OpenMP geometric multigrid application using hybrid parallelism.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Ground and excited state gradients with end-to-end differentiable semiempirical quantum chemistry

Accurate and efficient gradients of molecular energy with respect to nuclear degrees of freedom are essential for geometry optimization and molecular dynamics, including simulations that go beyond the Born–Oppenheimer regime. A common approach involves deriving analytical formulas for new electronic structure methods, which is often conceptually difficult and requires tedious coding. Here, we implement analytical, semi-numerical, and automatic differentiation (AD)-based gradient pathways for semiempirical Hamiltonian models in the PYSEQM software package, leveraging both graphics processing unit (GPU) and central processing unit (CPU) architectures. We further extend these capabilities to excited states calculated using the configuration interaction singles and time-dependent Hartree–Fock ansätze. We benchmark wall time, peak memory usage, and accuracy across three molecular families of varying chemical complexity, including systems of up to a thousand atoms. For ground-state simulations, analytical and AD gradients achieve near-identical GPU runtimes, while semi-numerical gradients are slower on GPU but remain competitive on CPU. For excited states, both analytical and custom AD approaches using implicit differentiation show similar performance and low memory requirements, whereas gradients with full AD are memory-limited. AD gradients match analytical ones in accuracy across all tested systems, aided by a quaternion-based diatomic frame rotation for two-center quantities that ensures smooth energy surfaces. Overall, automatic differentiation emerges as a practical alternative to analytical gradients in semiempirical quantum chemistry, offering high accuracy while allowing seamless integration in AI-driven workflows and popular packages, such as PyTorch and JAX. Our results provide actionable guidance for selecting optimal gradient strategies in large-scale ground- and excited-state molecular dynamics simulations.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

GPU-accelerated multitiered iterative phasing algorithm for fluctuation X-ray scattering

The multitiered iterative phasing (MTIP) algorithm is used to determine the biological structures of macromolecules from fluctuation scattering data. It is an iterative algorithm that reconstructs the electron density of the sample by matching the computed fluctuation X-ray scattering data to the external observations, and by simultaneously enforcing constraints in real and Fourier space. This paper presents the first ever MTIP algorithm acceleration efforts on contemporary graphics processing units (GPUs). The Compute Unified Device Architecture (CUDA) programming model is used to accelerate the MTIP algorithm on NVIDIA GPUs. The computational performance of the CUDA-based MTIP algorithm implementation outperforms the CPU-based version by an order of magnitude. Furthermore, the Heterogeneous-Compute Interface for Portability (HIP) runtime APIs are used to demonstrate portability by accelerating the MTIP algorithm across NVIDIA and AMD GPUs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

CUDO: closed-form universal dwell-time optimization for computer-controlled optical surfacing

Precision optical figuring demands fast and accurate dwell time optimization to reach nanometer- and sub-nanometer-level accuracy in next-generation optical systems. We introduce CUDO (closed-form universal dwell-time optimization), the first, to the best of our knowledge, unified closed-form analytical framework that supports both function-form and matrix-form dwell time models in computer-controlled optical surfacing (CCOS). In contrast to traditional methods, which rely on iterative optimization and hyperparameter tuning, our framework derives direct analytical solutions with no adjustable parameters. This approach unifies the solution principles of existing methods within a single mathematical model, delivering three key advantages: (1) accuracy on par with, or superior to, iterative solvers, (2) substantial reduction in computation time, and (3) numerical robustness. Comparative studies with prior art confirm that closed-form solutions achieve equivalent residual error while removing runtime bottlenecks. By simplifying the implementation and enabling real-time, scalable deployment, CUDO establishes a practical foundation for future deterministic fabrication of large-aperture and high-performance optics.

36 MATERIALS SCIENCE↗
DOE OSTI2020

LDMS-GPU: Lightweight Distributed Metric Service (LDMS) for NVIDIA GPGPUs

GPUs are now a fundamental accelerator for many high-performance computing applications. They are viewed by many as a technology facilitator for the surge in fields like machine learning and Convolutional Neural Networks. To deliver the best performance on a GPU, we need to create monitoring tools to ensure that we optimize the code to get the most performance and efficiency out of a GPU. Since NVIDIA GPUs are currently the most commonly implemented in HPC applications and systems, NVIDIA tools are the solution for performance monitoring. The Light-Weight Distributed Metric System (LDMS) at Sandia is an infrastructure widely adopted for large-scale systems and application monitoring. Sandia has developed CPU application monitoring capability within LDMS. Therefore, we chose to develop a GPU monitoring capability within the same framework. In this report, we discuss the current limitations in the NVIDIA monitoring tools, how we overcame such limitations, and present an overview of the tool we built to monitor GPU performance in LDMS and its capabilities. Also, we discuss our current validation results. Most of the performance counter results are the same in both vendor tools and our tool when using LDMS to collect these results. Furthermore, our tool provides these statistics during the entire runtime of the tool as a time series and not just aggregate statistics at the end of the application run. This allows the user to see the progress of the behavior of the applications during their lifetime.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

$\mathrm{PPT}$-Multicore: performance prediction of Open$\mathrm{MP}$ applications using reuse profiles and analytical modeling

In this report we present PPT-Multicore, an analytical model embedded in the Performance Prediction Toolkit (PPT) to predict parallel applications’ performance running on a multicore processor. PPT-Multicore builds upon our previous work towards a multicore cache model. We extract LLVM basic block labeled memory trace using an architecture-independent LLVM-based instrumentation tool only once in an application’s lifetime. The model uses the memory trace and other parameters from an instrumented sequentially executed binary. We use probabilistic and computationally efficient reuse profiles to predict the cache hit rates and runtimes of OpenMP programs’ parallel sections. We model Intel’s Broadwell, Haswell, and AMD’s Zen2 architectures and validate our framework using different applications from PolyBench and PARSEC benchmark suites. The results show that PPT-Multicore can predict cache hit rates with an overall average error rate of 1.23% while predicting the runtime with an error rate of 9.08%.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Direction-optimizing Label Propagation and its Application for Community Detection

Label Propagation is a machine learning algorithm typically used for classification. It has also been found to be an effective method for detecting communities in networks. It has two attractive features as a community detection method: it has nearly linear runtime and it requires no \textit{a priori} community information. We propose a new Direction Optimizing Label Propagation Algorithm (DOLPA) that relies on the use of {\em frontiers} and alternates between label {\em push} and label {\em pull} operations to enhance the performance of LPA. Specifically, DOLPA has parameters for tuning the processing order of vertices in a graph. This reduces the number of edges visited and improves the quality of solution. We apply DOLPA to community detection and present the design and implementation of the algorithm as well as its shared-memory parallelization using OpenMP. Empirically, we evaluate our algorithm using synthetic graphs as well as real-world networks. Compared with the state-of-art \textit{Parallel Label Propagation} algorithm, we achieve at least two times the F-Score while reducing the runtime by 50\% for synthetic graphs with overlapping communities. We also compare DOLPA against the state-of-art parallel implementation of the Louvain method using the same graphs and show that DOLPA achieves about three times the F-Score at 10\% the runtime. On real-world graphs, we get a speedup of up to $10\times$ using 64 threads.

Liu, X↗
DOE OSTI2021

A Scalable Semi-Implicit Barotropic Mode Solver for the MPAS-Ocean

A scalable semi-implicit barotropic mode solver for the ocean component of the model for prediction across scales has been implemented as a competitor to an existing explicit-subcycling scheme to allow faster and more stable simulations while not sacrificing accuracy. The semi-implicit solver adopts the pipelined preconditioned bi-conjugate gradient stabilization algorithm as an iterative solver in conjunction with the restricted additive Schwarz preconditioner that accelerates the convergence rate of the iterative solver. The preconditioner is constructed from a linearized barotropic system that also reorders the system for optimal performance, while the semi-implicit solver deals with the fully nonlinear barotropic system that requires reassembly of the coefficient matrix for every time step. Several numerical experiments, from simple one-dimensional tests to three-dimensional real-world tests, demonstrate that the semi-implicit solver has almost the same accuracy and better parallel scalability compared with the existing scheme while allowing faster and more stable simulations. Furthermore, the semi-implicit solver accelerates the barotropic mode up to 2.9 times faster than the existing scheme on 16,320 processors, leading to an overall runtime speedup of 1.9.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2015

Diagnosing the Causes and Severity of One-sided Message Contention

Two trends suggest network contention for one-sided messages is poised to become a performance problem that concerns application developers: an increased interest in one-sided programming models and a rising ratio of hardware threads to network injection bandwidth. Unfortunately, it is difficult to reason about network contention and one-sided messages because one-sided tasks can either decrease or increase contention. We present effective and portable techniques for diagnosing the causes and severity of one-sided message contention. To detect that a message is affected by contention, we maintain statistics representing instantaneous (non-local) network resource demand. Using lightweight measurement and modeling, we identify the portion of a message's latency that is due to contention and whether contention occurs at the initiator or target. We attribute these metrics to program statements in their full static and dynamic context. We characterize contention for an important computational chemistry benchmark on InfiniBand, Cray Aries, and IBM Blue Gene/Q interconnects. We pinpoint the sources of contention, estimate their severity, and show that when message delivery time deviates from an ideal model, there are other messages contending for the same network links. With a small change to the benchmark, we reduce contention up to 50% and improve total runtime as much as 20%.

Network Contention, Performance Analysis, Dynamic ↗
DOE OSTI2018

I/O load balancing for big data HPC applications

High Performance Computing (HPC) big data problems require efficient distributed storage systems. However, at scale, such storage systems often experience load imbalance and resource contention due to two factors: the bursty nature of scientific application I/O; and the complex I/O path that is without centralized arbitration and control. For example, the extant Lustre parallel file system-that supports many HPC centers-comprises numerous components connected via custom network topologies, and serves varying demands of a large number of users and applications. Consequently, some storage servers can be more loaded than others, which creates bottlenecks and reduces overall application I/O performance. Existing solutions typically focus on per application load balancing, and thus are not as effective given their lack of a global view of the system. In this paper, we propose a data-driven approach to load balance the I/O servers at scale, targeted at Lustre deployments. To this end, we design a global mapper on Lustre Metadata Server, which gathers runtime statistics from key storage components on the I/O path, and applies Markov chain modeling and a minimum-cost maximum-flow algorithm to decide where data should be placed. Evaluation using a realistic system simulator and a real setup shows that our approach yields better load balancing, which in turn can improve end-to-end performance.

Paul, Arnab K.↗