Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “SERVER”

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 1,045 records · Page 58

A Public-Private-Academic Partnership to Advance Solar Power Forecasting

Executive Summary NOAA is making major contributions to the solar forecasting project in three areas. First, it is improving its forecasts of solar irradiance, clouds, and aerosols in its numerical weather prediction models. Second, it is providing advanced satellite products for DOE's FOA awardees to use in their forecast systems. Third, it is using high-quality ground-based measurements from SURFRAD and ISIS stations to verify and validate forecast model output. This reports covers results from all three areas for the period May 1, 2014 - April 30, 2015. Modeling In its modeling effort, NOAA continues work to improve the skill of solar forecasts from the Earth System Research Lab (ESRL) research versions of the 13-km Rapid Refresh (RAP) and the 3-km High-Resolution Rapid Refresh (HRRR) models, which are in turn transitioned into operations at the National Centers for Environmental Prediction (NCEP). A major milestone was achieved in September 2014 with the initial operational implementation of the HRRR at NCEP. In the ESRL research versions of the models, testing and development, in both real-time runs and retrospective experiments, is guided by an extensive in-house verification system. Early in the SFIP project, we developed the capability to verify our model forecasts against the high-quality surface radiation measurements from the SURFRAD and ISIS networks. This highlighted some shortcomings with the RAP and HRRR forecasts of incoming shortwave radiation. Most of our effort during Phase 1 of SFIP was focused on addressing these problems with a variety of model system improvements. The RAP and HRRR models during the warm season of 2014 had a noticeable warm and dry bias in near-surface conditions over most of the central and eastern United States, and our new SURFRAD/ISIS verification revealed that there was also a large excess of incoming global horizontal irradiance in the models. We hypothesized that a lack of cloud cover (particularly low-level cloud cover) in the models was resulting in too much heating of the land surface. This, in turn, caused unrealistically strong surface heat fluxes and turbulent mixing in the planetary boundary layer (PBL), which further reduced the already deficient cloud cover. We addressed these issues with a combination of data assimilation system modifications and model physics improvements. Many of our data assimilation changes were made with a view towards improving the near-term representation of clouds and precipitation. One of these changes involved better accounting for regions of weak reflectivity in the RAP cloud / hydrometeor assimilation system, in order to improve the representation of light precipitation in the RAP initial conditions and provide more realistic initial cloud cover. Additional modifications more accurately accounted for radar beam blockage and data gaps (particularly in the western United States), which improves shorter lead times forecasts of clouds and precipitation. We have also tested the assimilation of new data sources within the RAP and the HRRR, including radar radial velocity data and surface mesonet observations. Within the HRRR, we have tested the cycling of the 3-km land surface fields to allow a higher-resolution treatment of land surface processes. In terms of model physics development for SFIP, we have implemented a shallow cumulus scheme within the RAP, and have made numerous improvements to the Mellor-Yamada-Nakanishi-Niino (MYNN) PBL scheme to address insufficient low-level cloud cover in the models. We have conducted tests incorporating the radiation effects of (parameterized) boundary-layer clouds within the modified MYNN PBL scheme (independent of the convective schemes). The Grell-Freitas-Olson shallow cumulus scheme has also been tested within the 3-km HRRR. Finally, we have also modified the RUC land surface model (LSM) treatment of the vegetation wilting point, reducing it to increase evapotranspiration and increase cloud cover in the boundary layer. All of these changes work in tandem to significantly improve the model forecasts of cloud cover, incoming shortwave radiation, and near-surface temperature and moisture. Satellite The role of NOAA/NESDIS in the Solar Forecasting Improvement Project is to provide Advanced Satellite Products (ASPs) for the two forecasting teams at NCAR and IBM. The ASPs are cloud, surface, and atmosphere products derived from geostationary satellite imagery at the highest possible spatial and temporal resolution - such quantities as cloud mask, cloud probability, cloud transmission, cloud top height, cloud top temperature, cloud effective particle size, etc. Ancillary data, such as elevation and numerical weather prediction fields are provided in the files at the same resolution as well. There are at this time 147 different variables in the ASP output, including quality flags and processing information. The main goals for Year 1 of the project were to implement an Advanced Satellite Products system for the use of the IBM and NCAR teams, begin validation, and make any needed changes based on feedback from the teams. ASP files are being produced every GOES Imager acquisition, which occur on a 15-30 minute schedule. Processing is done on a dedicated computer, with a turn-around time of 8-21 minutes from image acquisition to results available on ftp. Several helpful visualizations of the data are also created for users on web pages. Users have been provided with a document titled "User's Guide for 1km Cloud Products Derived from GOES Imager Data using CLAVR-x", which discusses the basics of the source imagery, the process by which it is turned into Advanced Satellite Products, and considerations users should make when using the data. Validation of selected variables from the older 4km version of the products was also included. Future work will concentrate on validation of the 1km products and improving the turn-around time, product variety, and product quality as needed. Ground Observations In the ground-based measurement effort, NOAA's main objectives are to provide high quality radiation products for validation and verification of short-term to day-ahead solar forecasts. More specifically for the three year project, our goals include (1) Maintaining and providing data from our 7 SURFRAD and 7 ISIS; (2) Update ISIS radiation measurements from 3 min to 1 min data: (3) Purchase and install new pyrheliometers for direct solar irradiance measurements at the 7 SURFRAD sites; (4) Building, testing, and deploying two mobile SURFRAD stations at two utility plants in collaboration with DOE sponsored partners, and includes ongoing maintenance and processing of the data at the mobile sites; (5) Upgrading the data acquisition and communications at 7 SURFRAD sites and 7 ISIS sites; (6) Providing radiation data at the 7 SURFRAD sites in near real-time; (7) Develop and provide aerosol optical depth and cloud images and cloud fraction at our two mobile sites; (8) Provide data recovery rates each year; (9) Provide temporally and spatially averaged radiation products for comparison to HRRR and RAP solar forecasts and advanced satellite products; (10) Provide a data-set for analysis of conversion of direct and diffuse to sloped surfaces; (11) and as time permits develop and provide spectral solar irradiance, cloud optical depth and spectral albedo from the mobile sites. Milestones this year include working with the DOE sponsored teams to find locations to deploy two mobile SURFRAD stations. One existing unit was deployed at a 30MW PV facility in the San Luis Valley in collaboration with Xcel and the NCAR team in August, 2014. The second unit was built and tested at our facilities in Boulder, CO and deployed near Green Mountain Power's Education Center in Rutland, VT in collaboration with Green Mountain Power and the IBM Team in October, 2014. Data processing was implemented and the radiation data from these two mobile sites have been made available on our ftp server in near real-time. We also are providing images and cloud fraction from the TSI cameras for these two mobile sites on our ftp site. Another milestone was upgrading our data acquisition and communication systems at 7 SURFRAD and 7 ISIS sites. We accelerated our schedule for these upgrades to provide timely radiation products. These upgrades allow more reliable and near-real time radiation data delivery to the DOE sponsored teams to meet their goals. Lastly, we changed the data rate at the ISIS sites from 3 min to 1 min.

14 SOLAR ENERGY↗

Energy Performance Evaluation of Aquila's Aquarius Fixed Cold Plate Cooling System at NREL's High Performance Computing Center

In the first half of 2018, as part of a partnership with Sandia National Laboratories, Aquila installed its fixed cold plate, liquid-cooled Aquarius rack solution for high-performance computing (HPC) clustering at the National Renewable Energy Laboratory's (NREL's) Energy Systems Integration Facility (ESIF). This new fixed cold plate, warm-water cooling technology together with a manifold design provides easy access to service nodes and eliminates the need for server auxiliary fans altogether. Aquila and Sandia National Laboratories chose NREL's HPC Data Center for the initial installation and evaluation because the data center is configured for liquid cooling and has the required instrumentation to measure flow and temperature differences to facilitate testing. This paper gives an overview of the Aquarius fixed cold plate cooling technology and provides results from early energy performance evaluation testing.

97 MATHEMATICS AND COMPUTING↗

Understanding the Hierarchy of Dense Subgraphs in Stationary and Temporally Varying Setting

Graphs are widely used to model relationships in a wide variety of domains such as sociology, bioinformatics, infrastructure, the WWW, to name a few. One of the key observations is that while real-world graphs are often globally sparse, they are locally dense. In other words, the average degree is often quite small (say at most 10 in a million vertex graph), but vertex neighborhoods are often dense. Finding dense subgraphs is a critical aspect of graph mining It has been used for finding communities and spam link farms in web graphs, graph visualization, real-time story identification, DNA motif detection in biological networks, finding correlated genes, epilepsy prediction, finding price value motifs in financial data, graph compression, distance query indexing, and increasing the throughput of social networking site servers. However, most standard formulations of this problem (like clique, quasi-clique, k-densest subgraph) are NP-hard. Furthermore, current dense subgraph finding algorithms usually optimize some objective, and only find a few such subgraphs without providing any structural relations, whereas the goal is rarely to find the "true optimum," but to identify many (if not all) dense substructures, understand their distribution in the graph, and ideally determine relationships among them. In this project, we first aim to devise algorithms and provide 3 implementations with nice visualizations to find the hierarchy between dense subgraphs, and then understand the structure of the hierarchy to gain more insight on the hidden patterns in real-world networks. Another important aspects in graph analysis is the temporal nature of networks. Networks evolve over time and in many applications data arrives at a high velocity, and thus it is important to design algorithms that can process data efficiently. We report three main results towards identifying dense structures in large evolving graphs. First, we will show how the hierarchical connectedness structure can be maintained efficiently, where connectedness is defined by increasing levels of connectivity strength. Next, we present dense structure can be identified in bipartite graphs without building projection graphs. And finally, we present a new method for peeling algorithms This new approach avoids sequential nature of peeling algorithms and is amenable to parallelization, which is crucial for processing high velocity data.

97 MATHEMATICS AND COMPUTING↗

Lazy Shadowing - An Adaptive, Power-Aware, Resiliency Framework for Extreme Scale Computing

The major goal of the project was to examine the problem of resilience in exascale systems via mechanisms that minimize the impact of server crashes resulting from both permanent faults and transient errors or subtle time-dependent bugs in hardware or software. Shadowing was proposed as the main technique for accomplishing this in the proposal, wherein some replicas of the application are run at much lower power and opportunistically speeded up during periods of low activity of the main or when the main fails. A related goal of the project was to detect the errors early so that program crashes could be avoided or the scope of the rollbacks could be limited.

97 MATHEMATICS AND COMPUTING↗

Open MPI Testing Infrastructure Update - September, 2019

Since the last update on the Open MPI regression testing infrastructure, and the deployment of the MTT Python client on more DOE HPC centers, problems with the client and its interaction with the community PostGreSQL database server at AWS began to appear. In particular, problems were found with the reporting of failed and timed out tests to the community database.

97 MATHEMATICS AND COMPUTING↗

Open V2X at ESIF: June 2015 - February 2017 (Final Report)

The University of Delaware (UD), and their industrial partners, have carried out testing, documentation, and standards development at ESIF. These activities are intended to help meet the objective of NREL's INTEGRATE program, to enable clean energy technologies to increase the hosting capacity of the grid, by providing grid services in a holistic manner using an open source, interoperable platform. In particular, the 'Open V2X at ESIF' project worked to facilitate standardization and interoperability for Electric Vehicle technologies, including battery electric vehicles and plug-in hybrid vehicles (hence EV) to provide grid services. This was done by interconnecting and testing EV and EVSE (charging station) hardware from different manufacturers, by developing standards and test protocols for standards, and by documenting new protocols to be proposed as standards. Contractor-supplied equipment used in the testing included Vehicle Smart Links (VSLs) installed in an EV and School Bus, single-phase and three-phase charging stations, and an optimizing aggregator (software running on a server, hence just 'Aggregator'). The Aggregator software was developed by U of Delaware and is now licensed to Nuvve Corp for commercial development, and licensed to NRG/EVgo for demonstrations. The Aggregator monitors capacity and operations of the EVs and EVSEs, dispatches them in response to requests, and reports their operations and metered power flow. In operation, as vehicles move, plug-in and unplug, the aggregator tracks the total capacity available within a defined jurisdiction or RTO (Regional Transmission Organization, synonym for ISO or TSO), and dispatches part or all of that capacity based on market or direct dispatch signals from a grid entity. The VSL is the data acquisition and control module in the resource itself (EVSE or EV), with rich communication between a single Optimizing Aggregator and many VSL (one in each resource). All requests and dispatch events are also logged by the aggregator to make the provision of grid power auditable. For these tests, the dispatch was typically a recorded signal, the VSL was in the EV or EVSE, and the recording of experimental performance was often carried out by the normal logging functions of the aggregator.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Dynamic Contingency Analysis Tool 2.0 User Manual with Test System Examples

This document introduces Version 2 of the Dynamic Contingency Analysis Tool (DCAT) software package developed by Pacific Northwest National Laboratory and serves as a guide for using this package to conduct cascading failure simulations for the electric power system. The DCAT is an open-platform and publicly available software; it is intended to help develop applications that aim to improve the capabilities of power system planning engineers to assess the impact and likelihood of extreme contingencies and potential cascading events across their systems and interconnections. Outputs from the DCAT will support finding mitigating actions to reduce the risk of cascading outages in technically sound and effective ways. This manual provides detailed instructions on how to use this tool on multicore Windows workstations or servers. In the public version of DCAT, several improvements have been made over the previous version: (1) it was modularized; (2) a graphical user interface (GUI) was developed; (3) a new version of the MPjobs parallel processing module was incorporated; and (4) a batch-processing function was added. The features of the DCAT GUI are described, including some examples of how to use the GUI to perform simulations. A full explanation of the DCAT methodology is available in the Phase 1 report.

24 POWER TRANSMISSION AND DISTRIBUTION↗

A Study of Geodesic Distance Kernel on an Integrated GPU

Graphics processing units (GPUs) are commonly used for graphics and general-purpose (GPGPU) computation. Currently, NVIDIA’s GPUs, through its CUDA framework for GPGPU programming, are ubiquitous for highperformance computing. Another class of GPUs, with a central processing unit (CPU) and a GPU integrated on the same chip, is commonly used in laptops, desktop computers, and low-power servers. While they are not designed to outperform discrete GPUs due to the power, area, and thermal constrains [1], there is a need to better understand the performance of a processor with an integrated GPU for floating-point intensive applications. The study helps us better understand the characteristics of hardware and software development tools, and the benefit of offloading such applications to an integrated GPU.

97 MATHEMATICS AND COMPUTING↗

Technical Evaluation Report for the NDMAS AGR-5/6/7 TRISO Irradiation Monitoring System

The Nuclear Data Management and Analysis System (NDMAS) manages irradiation experiment data collected under the Advanced Gas Reactor (AGR) Fuel Development and Qualification program within the Advanced Reactor Technologies Technology Development Office. The AGR-5/6/7 irradiation experiment combines the fifth, sixth, and seventh planned irradiation campaigns of the U.S. Department of Energy AGR Fuel Development and Qualification Program. These combined irradiation tests will consist of fuel qualification tests (AGR-5/6) and a margin test (AGR-7) for the industrially produced uranium oxycarbide (UCO) tristructural isotropic (TRISO) coated particle fuel developed by the AGR program. The objective of the AGR-5/6/7 irradiation experiment is to establish the performance of UCO TRISO fuel fabricated using engineering-scale equipment in a production environment under and beyond normal operating conditions (PNL-5245). Experimental irradiation measurements for this experiment (i.e., thermocouple temperature, gas flow rate, fission product release rate, gross gamma count, and Advanced Test Reactor (ATR) operational parameters) are captured automatically to the NDMAS database. Data derived from numerical simulation or other mathematical analyses (i.e., fast fluence, burnup, temperature, and fission product release to birth ratio) are captured to the database when provided by the researchers. The NDMAS database is a relational database accessed via structured query language and managed with Microsoft SQL Server. While select users can directly access the database, that is not the most convenient way to interact with the data. The NDMAS team constructs numerous tables that combine related datasets, and delivers those data to researchers via graphs and spreadsheets on a Microsoft SharePoint-based website. For the AGR-5/6/7 experimental data, a new subsite has been developed, along with a set of web pages. These pages will be used by members of fuel irradiation team for monitoring irradiation conditions according to specification (SPC-1749), by the physics modelers to access necessary model inputs, and by the nuclear engineering community to view and access data such as fuel temperatures, fission product release from capsules, and ATR operational parameters. The deliverable represented in this document is a SharePoint website containing tables of the AGR-5/6/7 measured and calculated data, graphs displaying the data, and links to source documents and references.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

INTEGRATED WORKFLOW MANAGEMENT FOR PARTICLE ACCELERATOR SIMULATION

Supercomputing systems are used for a wide range of computationally demanding tasks in many fields of science and engineering. They play a key role in numerical simulation, in which mathematical models are computed in order to simulate the behavior of physical systems. Scientists and engineers that use supercomputers for numerical simulation often have their productivity limited by the need to manually organize and manage extremely large amounts of data that are often produced and consumed by the software programs run on these systems. Recognizing these limitations, Kitware Inc. (Clifton Park, NY) and SLAC National Accelerator Laboratory (Menlo Park, CA) are developing an advanced software platform that can reduce the cognitive overhead required by knowledge workers when using supercomputers for numerical simulation. Phase I of the project is complete and includes the development of new capabilities for organizing simulation project files, improvements to the user interface and overall usability, and deployment of a “middle tier” server to sit between user desktop machines and supercomputers to offload much of the data management workload. The project also developed prototype software for executing sequences of numerical simulations, and a prototype for migrating supercomputing software to cloud-based computing systems to provide a potential alternative to supercomputers with different logistical and price-to-performance tradeoffs.

Tourtellott, John↗

Data Analytics for Electrical Distribution Systems with Micro PMUs (GMLC 1.4.9 Technical Report)

This report describes how electrical utilities with residential, industrial, commercial and government customers make decisions using Data Analytics (DA) applications and discusses how OpenFMB was leveraged to share that data among disparate types of devices for DA. OpenFMB is a data standard developed by utilities for distributed communication and control of power system assets. Here we explored how OpenFMB can be used for sharing analytics and what those analytics could be for a set of representative grid devices. A review of DA applications performed by electrical utilities and customers is presented. An overview of how the testbed we used was built and configured is summarized. Several experiments which were carried out to explore the capabilities of OpenFMB alongside other data sharing mechanisms for gathering sensor measurements and sharing analytics is also discussed. And finally a comparison of OpenFMB and LBNL’s SPARK framework is summarized. Today’s electrical utilities focus on monitoring the real-time customer consumptions, offering new real-time tariff models based on customer usage habits (rewards and punishments), providing energy peak during peak load times (plan energy generation and distribution), improving the load management for transformers and power lines, and controlling the energy and grid resources efficiently. On the other hand, the DA applications for residential customers are based on modifying the customer electricity usage habit to save money, making accurate and timely decisions for electrical billing. However, the DA applications for industrial, commercial and government customers focus on creating efficient programs to improve the environmental condition (reducing CO₂ emission), developing better load forecasting for industrial and/or commercial activities. In this project, a testbed was built to explore how the OpenFMB framework can be used to distribute data among various types of representative grid edge devices. Micro PMUs (Phasor Measurement Unit) were used to collect and transmit data to an OpenPDC server alongside relays with a framework configured with a system using OpenFMB to collect and transmit data. The testbed was composed of a physical asset based on a 60 Hz AC radial power system constructed using SI-GRID components. The power system consisted of a 3-phase 14 volts line-to-ground (24 volts line-to-line) power source, three 3-phase line emulators, a programmable resistive load bank, two smart relays, two micro PMUs, and a (SCADA) and Phasor Data Concentrator (PDC) computer. This testbed configuration allowed us to perform breaker operation tests collecting data from the micro PMUs and from the relays’ sensors. During the micro PMU breaker operation tests we collected the measured and estimated fundamental voltages, currents, frequency as well as the total true, reactive and apparent power. The displacement power factor from two micro PMUs located in parallel power lines was also used. The software steps for the implementation of the Synchrophasor system are explained in detail in this report. The setting (1-2-3) and data collection (4-5) phases were performed with a free download open source software. In the setting phase, the micro PMUs were set (1), the data stream for each micro PMU was validated, and the connection and configuration files were created (2). The PDC computer was set using the connection and configuration files of the micro PMUs (3). In the data collection phase, the binary data from the micro PMUs was collected (4), and the binary data files of the micro PMUs was converted to analyze the test results (5). OpenFMB was configured to use protocol translators to read Modbus and transform the reading into the OpenFMB data model to be read for DA applications.Our experiments showed that by leveraging the OpenFMB data model and the micro PMUs together we could obtain better situational awareness of network faults, which in some cases could lead to failures in the system and would be less perceivable otherwise. For example, in one experiment the fundamental RMS current angle, and the fundamental displacement power factor (estimated by the ratio between the measured real and apparent total power) had a disturbance at the micro PMU connected to the beginning of the parallel power line because the breaker was opened without the operators or SCADA system’s knowledge, and no current was available. Based on the breaker state scenarios and micro PMU locations, the DA applications using the current angle and displacement power factor from the micro PMUs should be studied in detail to avoid a non-desired measurement that could affect the decision from the DA. Also, the sample rates for PMU, Relay and micro PMU devices were compared. The DA application improved the control and operation between the electrical utilities and their customers (residential, industrial, transportation, commercial and government). The higher the sample rate, the better the quality of the measurement for DA estimations.

42 ENGINEERING↗

Drone Video Platform—Collision Avoidance, Situational Awareness, and Communications

Small unmanned aerial systems are ideal for delivering sensors into areas that are inaccessible or too dangerous for human entry or manned aircraft over-flights. The systems can fly low and slow for enhanced sensitivity and in tight spaces where ordinary aircraft could never fit. We have configured and flown sensor payloads for radiation, chemical, and optical detection on small fixed-wing and rotary-wing aircraft. All the sensor packages provide telemetry, GPS coordinates, LIDAR distance ranging, and absolute time of day to microsecond accuracy. Payload weights range from less than 1 kg up to 5.5 kg. Currently, communications are over 2.4 GHz and 915 MHz radios, as well as over 4G LTE cellular data links. Real-time data feeds from the sensor payloads to web browser clients anywhere in the world are possible via a virtual EC2 server on the Amazon AWS cloud. In FY 2019, the Nevada National Security Site teamed with Virginia Tech, H3D Corp., and Unmanned Systems, Inc. to accomplish the four flight missions described in this report. A large part of this year’s mission goals were to demonstrate the ability to fly beyond visual line of sight (BVLOS).

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Integrating System to Edge-of-Network Architecture and Management for SHINES (SEAMS) Technologies of High Penetration Grids

Consistent with the U.S. DOE’s EERE SHINES FOA objectives, the goal of this project was to enable integration from EMS to variable DER in a way that delivered visibility and opportunities for managing and controlling distributed resources using SEAMS technologies. This project implemented several SHINES technologies, which included: (1) three Stem battery systems; (2) six ConnectDER PV collars; (3) two E-Gear PV/Battery storage interfaces; and (4) eight Kitu smart inverter emulators. A Siemens Energy IP DEMS was deployed in a development environment and integrated with the SHINES technologies via OpenADR and IEEE 2030.5 protocols with documented architecture. In the case of IEEE 2030.5, a suite of technologies including a gateway, an API, software and a server from Kitu Systems (the Citadel, Convoy, Kitu Crosslink API, and Kitu Spark) were used to demonstrate that the utility could achieve visibility of distributed energy resources (DER) and control of single and multiple DER systems with a single command. The project also included a LCOE analysis of SHINES technologies compared to traditional system and circuit mitigation measures.

14 SOLAR ENERGY↗

Big Data Processing for Power Grid Event Detection

In this paper we present the application of big data processing for the development of machine learning(ML) models to detect relevant events in power gridoperations. This is based on almost 20TB of phasormeasurement unit data corresponding to up to two years of operation of three grid interconnections which provide power to most of the United States. A significant aspect of the work consists in having all data processing performed on a single standard GPU server, from pre-processing to ML model training and testing. We describe the data and computational infrastructure, challenges faced and methods used in dataprocessing, main findings and results. The ML approach employed for best utilization of the big data is also discussed, including sample results.

Paes Leao, Bruno↗

Understanding HPC Benchmark Performance on Intel Broadwell and Cascade Lake Processors

Hardware platforms in high performance computing are constantly getting more complex to handle even when considering multicore CPUs alone. Numerous features and configuration options in the hardware and the software environment that are relevant for performance are not even known to most application users or developers. Microbenchmarks, i.e., simple codes that fathom a particular aspect of the hardware, can help to shed light on such issues, but only if they are well understood and if the results can be reconciled with known facts or performance models. The insight gained from microbenchmarks may then be applied to real applications for performance analysis or optimization. In this paper we investigate two modern Intel x86 server CPU architectures in depth: Broadwell EP and Cascade Lake SP. We highlight relevant hardware configuration settings that can have a decisive impact on code performance and show how to properly measure on-chip and off-chip data transfer bandwidths. The new victim L3 cache of Cascade Lake and its advanced replacement policy receive due attention. Finally we use DGEMM, sparse matrix-vector multiplication, and the HPCG benchmark to make a connection to relevant application scenarios.

97 MATHEMATICS AND COMPUTING↗