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 829 records · Page 46

FPDeep: Scalable Acceleration of CNN Training on Deeply-Pipelined FPGA Clusters

In this paper, we propose a framework called FPDeep, which uses a hybrid of model and layer paral- lelism to configure distributed reconfigurable clusters to train DNNs. This approach has numerous benefits. First, the design does not suffer from batch size growth. Second, novel workload and weight partitioning leads to balanced loads of both among nodes. And third, the entire system is fine-grained pipeline. This leads to high parallelism and utilization and also minimizes the time features need to be cached while waiting for back-propagation.

Wang, Tianqi↗

Enforcing End-to-end I/O Policies for Scientific Workflows using Software-Defined Storage Resource Enclaves

Data-intensive knowledge discovery requires scientific applications to run concurrently with analytics and visualization codes executing in situ for timely output inspection and knowledge extraction. Consequently, I/O pipelines of scientific workflows can be long and complex because they comprise many stages of analytics across different layers of the I/O stack of high-performance computing systems. Performance limitations at any I/O layer or stage can cause an I/O bottleneck resulting in greater than expected end-to-end I/O latency. In this paper, we present the design and implementation of a novel data management infrastructure called Software-Defined Storage Resource Enclaves (SIREN) at system level to enforce end-to-end policies that dictate an I/O pipeline's performance. SIREN provides an I/O performance interface for users to specify the desired storage resources in the context of in-situ analytics. If suboptimal performance of analytics is caused by an I/O bottleneck when data are transferred between simulations and analytics, schedulers in different layers of the I/O stack automatically provide the guaranteed lower bounds on I/O throughput. Lastly, our experimental results demonstrate that SIREN provides performance isolation among scientific workflows sharing multiple storage servers across two I/O layers (burst buffer and parallel file systems) while maintaining high system scalability and resource utilization.

97 MATHEMATICS AND COMPUTING↗

FELIX-Based Readout of the Single-Phase ProtoDUNE Detector

Large liquid argon (LAr) time projection chambers (TPCs) have been adopted for the Deep Underground Neutrino Experiment (DUNE) experiment's far detector, which will be composed of four 17-kton detectors situated 1.5 km underground at the Sanford Underground Research Facility. This represents a large increase in scale compared to existing experiments. Both single- and dual-phase technologies will be validated at CERN, in cryostats capable of accommodating full-size detector modules, and exposed to low-energy charged particle beams. This program, called ProtoDUNE, also allows for extensive tests of data acquisition strategies. The Front-End LInk eXchange (FELIX) readout system was initially developed within the ATLAS collaboration and is based on custom field-programmable gate array (FPGA)-based Peripheral Component Interconnect Express input/output cards, connected through point-to-point links to the detector front end and hosted in commodity servers. FELIX will be used in the single-phase ProtoDUNE setup to read the data coming from 2560 anode wires organized in a single anode plane assembly (APS) structure. With a continuous readout at a sampling rate of 2 MHz, the system must deal with an input rate of 96 Gb/s. An external trigger will preselect time windows of 5 ms with interesting activity expected inside the detector. Event building will occur for triggered events, at a target rate of 25 Hz; the readout system will form fragments from the data samples matching the time window, carry out lossless compression, and forward the data to event building nodes over 10-Gb/s Ethernet. This paper discusses the design and implementation of this readout system as well as the first operational experience.

47 OTHER INSTRUMENTATION↗

Machine Learning-Based Extreme Data Reduction for Prompt Supernova Pointing at DUNE

One of the goals of the Deep Underground Neutrino Experiment (DUNE) is to use the massive underground liquid argon time projection chamber (LArTPC) detectors at its far site for multimessenger astronomy (MMA), in the detection of neutrinos from core-collapse supernovae (SNe). Its current baseline trigger strategy detects activity in the detector that is consistent with supernova (SN) neutrinos and saves the raw data for further offline analysis but provides no prompt pointing information crucial for optical follow-ups by other observatories. This approach is based on the assumption that prompt pointing determination using raw data is computationally prohibitive. In this article, we demonstrate a proof-of-concept based on applying extreme data reduction on the buffered SN data in the DUNE data acquisition (DAQ) system’s front-end computers using a machine learning (ML) workflow. This reduces the data by ~5 orders of magnitude, allowing a full track reconstruction to be carried out quickly on a single server. The total time to perform the ML-based data reduction and the full track reconstruction is less than the time to transfer the SN data back to Fermilab or a high-performance computing (HPC) center. This shows that prompt processing of raw SN data is possible and, in fact, trivial once the data have been reduced to reject radiological backgrounds, paving the way to a high-quality SN pointing trigger that is based on fully reconstructed data instead of trigger primitives (TPs).

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Distributed Finite-Time Termination for Consensus Algorithm in Switching Topologies

Here, in this article, we present a finite-time stopping criterion for consensus algorithms in networks with dynamic communication topology. Prior state of the art has established convergence to the consensus value; however, the asymptotic convergence of these algorithms poses a challenge in practical settings where the response from agents is required in finite time. To this end, we propose a maximum-minimum protocol that propagates the global maximum and minimum values of agent states (while running the consensus algorithm) in the network. This article focuses on establishing that the global maximum and minimum values are strictly monotonic even for a dynamic topology, and they can be used to distributively ascertain the closeness to convergence in finite time. We rigorously show that each node can have access to the global maximum and minimum by running the proposed maximum-minimum protocol to realize a finite-time stopping criterion for the otherwise asymptotic consensus algorithm. The practical utility of the algorithm is illustrated through experiments where each agent is instantiated by a NodeJS socket.io server.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Evaluating Modern GPU Interconnect: PCIe, NVLink, NV-SLI, NVSwitch and GPUDirect

High performance multi-GPU computing becomes an inevitable trend due to the ever-increasing demand on computation capability in emerging domains such as deep learning, big data and planet-scale simulations. However, the lack of deep understanding on how modern GPUs can be connected and the real impact of state-of-the-art interconnect technology on multi-GPU application performance become a hurdle. In this paper, we fill the gap by conducting a thorough evaluation on five latest types of modern GPU interconnects: PCIe, NVLink-V1, NVLink-V2, NVLink-SLI and NVSwitch, from six high-end servers and HPC platforms: NVIDIA P100-DGX-1, V100-DGX-1, DGX-2, OLCF’s SummitDev and Summit supercomputers, as well as an SLI-linked system with two NVIDIA Turing RTX-2080 GPUs. Based on the empirical evaluation, we have observed four new types of GPU communication network NUMA effects: three are triggered by NVLink’s topology, connectivity and routing, while one is caused by PCIe chipset design issue. These observations indicate that, for a multi-GPU application running in a multi-GPU node, choosing the right GPU combination can impose considerable impact on GPU communication efficiency, as well as an application’s overall performance. Our evaluation can be leveraged in building practical multi-GPU performance models, which are vital for GPU task allocation, scheduling and migration in a shared environment (e.g., AI cloud and HPC centers), as well as communication-oriented performance tuning.

Performance Evaluation, GPU, Interconnect, NUMA, P↗

Customizable Scale-Out Key-Value Stores

Enterprise KV stores are often not well suited for HPC applications, and thus cumbersome end-to-end KV design customization is required to meet the needs of modern HPC applications. To this end, in this article we present bespoKV, an adaptive, extensible, and scale-out KV store framework. bespoKV decouples the KV store design into the control plane for distributed management and the data plane for local data store. For the control plane, bespoKVprovides pre-built modules, called controlets, supporting common distributed functionalities (e.g., replication, consistency, and topology) and their various combinations. This decoupling allows bespoKV to take a user-provided single-server KV store, called a datalet, and transparently enables a scalable and fault-tolerant distributed KV store service. The resulting distributed stores are also adaptive to consistency or topology requirement changes and can be easily extended for new types of services. Such specializations enable innovative uses of KV stores in HPC applications, especially for emerging applications that utilize KV-friendly workloads. We evaluate bespoKV in a local testbed as well as in a public cloud settings. Experiments show that bespoKV-enabled distributed KV stores scale horizontally to a large number of nodes, and performs comparably and sometimes 1.2× to 2.6× better than the state-of-the-art systems.

97 MATHEMATICS AND COMPUTING↗

An Integrated Indexing and Search Service for Distributed File Systems

Data services such as search, discovery, and management in scalable distributed environments have traditionally been decoupled from the underlying file systems, and are often deployed using external databases and indexing services. However, modern data production rates, looming data movement costs, and the lack of metadata, entail revisiting the decoupled file system-data services design philosophy. In this article, we present TagIt, a scalable data management service framework aimed at scientific datasets, which can be integrated into prevalent distributed file system architectures. A key feature of TagIt is a scalable, distributed metadata indexing framework, which facilitates a flexible tagging capability to support data discovery. Furthermore, the tags can also be associated with an active operator, for pre-processing, filtering, or automatic metadata extraction, which we seamlessly offload to file servers in a load-aware fashion. We have integrated TagIt into two popular distributed file systems, i.e., GlusterFS and CephFS. Our evaluation demonstrates that TagIt can expedite data search operation by up to 10× over the extant decoupled approach.

97 MATHEMATICS AND COMPUTING↗

Impact of GPS Signal Loss and Its Mitigation in Power System Synchronized Measurement Devices

With the aid of global positioning system (GPS), synchronized measurement devices (SMDs) are increasingly deployed across power systems to monitor the status of electric grids by providing accurate measurement data along with unified time stamps. Unfortunately, GPS receivers tend to lose signal lock when certain uncontrollable and unpredictable factors arise. In order to investigate the presence of GPS signal loss (GSL) issues on measurement devices, analysis is performed on historical data from both phasor data concentrators and FNET/GridEye servers. Meanwhile, the impact of GSL on field measurement accuracy has not been previously explored in depth. Through analysis and experimental tests, this paper discovers angle drift caused by GSL, which consequently leads to the total vector error exceeding the IEEE standard C37.118.1-2011. Furthermore, a compensation method is proposed to rectify the angle drift and laboratory experiments demonstrate that the proposed method does effectively reduce angle drift and mitigate the impact of GSL in SMDs.

47 OTHER INSTRUMENTATION↗

UPS: Unified PMU-Data Storage System to Enhance T+D PMU Data Usability

The emerging distribution-level phasor measurement unit (D-PMU) is expected to play an important role in enhancing distribution system observability and situational-awareness. It is a demanding yet challenging task to develop advanced D-PMU data management and analytics tools to improve D-PMU data usability and further promote D-PMU projects deployment. This paper focuses on D-PMU data processing and storage. It presents a brief review of existing D-PMU data storage systems and points out their limitations on high-performance, flexibility, and scalability. To overcome the limitations, a unified PMU-data storage system (UPS) is proposed. Specifically, a unified I/O interface between storage servers and computing jobs is developed to effectively reduce the overhead of managing various computing jobs and data analytics over multiple storage infrastructures; and PMUCache with PMUCache partitioning and PMUCache replacement algorithms are designed to support in-situ data processing and shared distributed data storage and further serve the computing jobs/queries with different Quality of Service (QoS) requirements. Through a series of experiments, it is demonstrated that UPS achieves high performance on fast and big data processing and storage, and efficiently increases the flexibility and scalability of PMU data management systems.

42 ENGINEERING↗

Secure mmWave Spectrum Sharing with Autonomous Beam Scheduling for 5G and Beyond

Spectrum Sharing (SS) has seen a renewed set of initiatives in 5G with the availability of shared and unlicensed spectrum bands that can be used by multiple cellular service providers and private cellular networks. Beam based transmission, instead of the traditional sector based transmission in conjunction with the spectrum agility of the 5G New Radio (NR) has brought new opportunities to optimized sharing of spectrum. Currently in the U.S., a centralized Spectrum Access Server (SAS) is used to co-ordinate spectrum sharing among networks sharing the same spectrum band. However, SAS becomes a focal point for security attacks and a performance bottleneck. In addition, SAS relies on an Environmental Sensor Network (ESN), separate from the 5G network. Without trusted spectral occupancy information, false reporting of spectrum sensing data can create sub-optimal and unfair spectrum usage. This paper summarizes our recent research findings in using a decentralized scheme for multiple networks to securely share spectrum with autonomous beam scheduling : 1) A new stochastic network framework based on Lyapunov Optimization approach is developed to optimize scheduling at the base stations; 2) Game theoretic (GT) approach is used to formulate the distributed scheduler; 3) Another distributed scheduler with Q-learning is presented that utilizes the Reinforcement Learning (RL) approach; 4) The performance and convergence rate of these distributed solutions to use shared and unlicensed spectrum are compared with existing solutions. Conditions under which the performance of these schedulers approach the theoretical upper bound, which is the performance possible with no interference among the operators sharing the spectrum, are presented; 5) The ability of a base station to use its own user equipment as sensors, for optimal spectrum sharing with base stations in other operator networks, is demonstrated to be an effective approach.

5G↗

Constraint-based modelling captures the metabolic versatility of Desulfovibrio vulgaris

A refined Desulfovibrio vulgaris Hildenborough flux balance analysis (FBA) model (iJF744) was developed, incorporating 1016 reactions that include 744 genes and 951 metabolites. A draft model was first developed through automatic model reconstruction using the ModelSeed Server and then curated based on existing literature. The curated model was further refined by incorporating three recently proposed redox reactions involving the Hdr-Flx and Qmo complexes and a lactate dehydrogenase (LdhAB, DVU 3027-3028) indicated by mutation and transcript analyses to serve electron transfer reactions central to syntrophic and respiratory growth. Eight different variations of this model were evaluated by comparing model predictions to experimental data determined for four different growth conditions -three for sulfate respiration (with lactate, pyruvate or H2/CO2-acetate) and one for fermentation in syntrophic coculture. The final general model supports (i) a role for Hdr-Flx in the oxidation of DsrC and ferredoxin, and reduction of NAD+ in a flavin-based electron confurcating reaction sequence, (ii) a function of the Qmo complex in receiving electrons from the menaquinone pool and potentially from ferredoxin to reduce APS and (iii) a reduction of the soluble DsrC by LdhAB and a function of DsrC in electron transfer reactions other than sulfite reduction.

59 BASIC BIOLOGICAL SCIENCES↗

Special Section on InterPACK 2017—Part 1

InterPACK is a premier international forum for exchange of state-of-the-art knowledge in research, development, manufacturing, and applications of micro-electronics packaging. It is the flagship conference of the ASME Electronic and Photonic Packaging Division (EPPD) founded in 1992 as an ASME-JSME joint biannual conference. Rapid changes in the semiconductor landscape together with findings from InterPACK Pathfinding workshop (IPW) in 2016 led to a significant reset of InterPACK conference priorities and focus to comprehensively address needs of the InterPACK community. As a result, starting in 2017, InterPACK has become an annual conference and the scope of the conference has increased significantly together with a systems-focus to include some of the most cutting-edge topics in electronics packaging, device integration, and reliability. These topics are organized across five different tracks: (1) heterogeneous integration: microsystems with diverse functionality, (2) servers of the future, (3) structural and physical health monitoring, (4) energy conversion and storage, and (5) transportation: autonomous and electric vehicles.

3D-printing↗

CPU and memory efficient coherent mode decomposition for the partially coherent x-ray simulations

In this work, the method of the Coherent Mode Decomposition (CMD) is applied to numerical wave propagation calculations for partially-coherent X-rays, using the Fourier optics and compatible methods. Its CPU and memory efficiency is discussed in various cases of the wavefront at the source and the beam waist. With the absence of the quadratic phase terms, the required sampling density of the electric fields is effectively reduced. The problem size is thus moderate and the method is feasible to be implemented on a single-node CPU server. In other cases, the same argument holds with proper treatments of the quadratic phase terms. Tests on CMD and the modes propagation are done for the case of the Coherent Hard X-ray beamline of the National Synchrotron Light Source II, using the Synchrotron Radiation Workshop software. We observe a few hundred or less dominant decomposed modes that resemble the electric fields converge to the wavefront intensity at a high accuracy of over 99%.

36 MATERIALS SCIENCE↗

Automated size-specific CT dose monitoring program: Assessing variability in CT dose

Purpose: The potential health risks associated with low levels of ionizing radiation have created a movement in the radiology community to optimize computed tomography (CT) imaging protocols to use the lowest radiation dose possible without compromising the diagnostic usefulness of the images. Despite efforts to use appropriate and consistent radiation doses, studies suggest that a great deal of variability in radiation dose exists both within and between institutions for CT imaging. In this context, the authors have developed an automated size-specific radiation dose monitoring program for CT and used this program to assess variability in size-adjusted effective dose from CT imaging. Methods: The authors radiation dose monitoring program operates on an independent health insurance portability and accountability act compliant dosimetry server. Digital imaging and communication in medicine routing software is used to isolate dose report screen captures and scout images for all incoming CT studies. Effective dose conversion factors (k-factors) are determined based on the protocol and optical character recognition is used to extract the CT dose index and dose-length product. The patient's thickness is obtained by applying an adaptive thresholding algorithm to the scout images and is used to calculate the size-adjusted effective dose (ED{sub adj}). The radiation dose monitoring program was used to collect data on 6351 CT studies from three scanner models (GE Lightspeed Pro 16, GE Lightspeed VCT, and GE Definition CT750 HD) and two institutions over a one-month period and to analyze the variability in ED{sub adj} between scanner models and across institutions. Results: No significant difference was found between computer measurements of patient thickness and observer measurements (p= 0.17), and the average difference between the two methods was less than 4%. Applying the size correction resulted in ED{sub adj} that differed by up to 44% from effective dose estimates that were not adjusted by patient size. Additionally, considerable differences were noted in ED{sub adj} distributions between scanners, with scanners employing iterative reconstruction exhibiting significantly lower ED{sub adj} (range: 9%-64%). Finally, a significant difference (up to 59%) in ED{sub adj} distributions was observed between institutions, indicating the potential for dose reduction. Conclusions: The authors developed a robust automated size-specific radiation dose monitoring program for CT. Using this program, significant differences in ED{sub adj} were observed between scanner models and across institutions. This new dose monitoring program offers a unique tool for improving quality assurance and standardization both within and across institutions.

60 APPLIED LIFE SCIENCES↗

SU-E-J-72: Geant4 Simulations of Spot-Scanned Proton Beam Treatment Plans

Purpose: To evaluate density inhomogeneities which can effect dose distributions for real-time image gated spot-scanning proton therapy (RGPT), a dose calculation system, using treatment planning system VQA (Hitachi Ltd., Tokyo) spot position data, was developed based on Geant4. Methods: A Geant4 application was developed to simulate spot-scanned proton beams at Hokkaido University Hospital. A CT scan (0.98 × 0.98 × 1.25 mm) was performed for prostate cancer treatment with three or four inserted gold markers (diameter 1.5 mm, volume 1.77 mm3) in or near the target tumor. The CT data was read into VQA. A spot scanning plan was generated and exported to text files, specifying the beam energy and position of each spot. The text files were converted and read into our Geant4-based software. The spot position was converted into steering magnet field strength (in Tesla) for our beam nozzle. Individual protons were tracked from the vacuum chamber, through the helium chamber, steering magnets, dose monitors, etc., in a straight, horizontal line. The patient CT data was converted into materials with variable density and placed in a parametrized volume at the isocenter. Gold fiducial markers were represented in the CT data by two adjacent voxels (volume 2.38 mm3). 600,000 proton histories were tracked for each target spot. As one beam contained about 1,000 spots, approximately 600 million histories were recorded for each beam on a blade server. Two plans were considered: two beam horizontal opposed (90 and 270 degree) and three beam (0, 90 and 270 degree). Results: We are able to convert spot scanning plans from VQA and simulate them with our Geant4-based code. Our system can be used to evaluate the effect of dose reduction caused by gold markers used for RGPT. Conclusion: Our Geant4 application is able to calculate dose distributions for spot scanned proton therapy.

07 ISOTOPE AND RADIATION SOURCES↗

SU-E-J-129: A Strategy to Consolidate the Image Database of a VERO Unit Into a Radiotherapy Management System

Purpose: To present a strategy to integrate the imaging database of a VERO unit with a treatment management system (TMS) to improve clinical workflow and consolidate image data to facilitate clinical quality control and documentation. Methods: A VERO unit is equipped with both kV and MV imaging capabilities for IGRT treatments. It has its own imaging database behind a firewall. It has been a challenge to transfer images on this unit to a TMS in a radiation therapy clinic so that registered images can be reviewed remotely with an approval or rejection record. In this study, a software system, iPump-VERO, was developed to connect VERO and a TMS in our clinic. The patient database folder on the VERO unit was mapped to a read-only folder on a file server outside VERO firewall. The application runs on a regular computer with the read access to the patient database folder. It finds the latest registered images and fuses them in one of six predefined patterns before sends them via DICOM connection to the TMS. The residual image registration errors will be overlaid on the fused image to facilitate image review. Results: The fused images of either registered kV planar images or CBCT images are fully DICOM compatible. A sentinel module is built to sense new registered images with negligible computing resources from the VERO ExacTrac imaging computer. It takes a few seconds to fuse registered images and send them to the TMS. The whole process is automated without any human intervention. Conclusion: Transferring images in DICOM connection is the easiest way to consolidate images of various sources in your TMS. Technically the attending does not have to go to the VERO treatment console to review image registration prior delivery. It is a useful tool for a busy clinic with a VERO unit.

60 APPLIED LIFE SCIENCES↗

SU-E-T-167: Evaluation of Mobius Dose Calculation Engine Using Out of the Box Preconfigured Beam Data

Purpose: Determine the dose calculation accuracy of a preconfigured Mobius server for use in secondary checks of a treatment planning system. Methods: 10 plans were created for irradiation on two of the IROC (formerly RPC) accreditation phantoms: 4 for the head and neck phantom and 6 for the lung phantom. The plans each were created using one of four different photon energies (6FFF, 10 FFF, 6X, and 15X) and were varied in treatment type including VMAT, step and shoot IMRT, dynamic MLC IMRT (DMLC), and conformal RT (CRT). The TLDs in the phantoms were contoured, and each plan was sent for calculation to Mobius software (Mobius Medical Systems, Houston, TX) with a default configuration. Each plan was then irradiated on the planned phantom 3 times to create an average reading across 56 TLDs. These readings were then compared against the corresponding Mobius calculation at each TLD location. Results: The mean difference (MD) normalized to the plan prescription dose between each TLD and Mobius calculation for all measurements was 0.5 ± 3.3%, with a maximum difference of 8.4%. The MD was 0.6 ± 3.8%, − 2.0 ± 1.9%, 1.7 ± 3.7%, and 1.9 ± 1.2% across the 6FFF, 10FFF, 6X and 15X energies respectively. The MD was −1.2 ± 2.3% for lung plans and 1.8 ± 3.5% for head/neck plans. Across treatment types, the MD ranged from − 1.8 ± 1.7% for CRT to 4.3 ± 2.4 % for DMLC. Conclusion: Out of the box and preconfigured, Mobius provides accurate dose calculations with respect to beam energy, treatment type, and treatment site.

60 APPLIED LIFE SCIENCES↗