检索结果-内蒙古大学图书馆

Horus: Interference-Aware and Prediction-Based Scheduling in Deep Learning systems

IEEE TRANSACTIONS ON parallel AND distributed systems 2022年第1期33卷 88-100页

作者： Yeung, Gingfung Borowiec, Damian Yang, Renyu Friday, Adrian Harper, Richard Garraghan, Peter Univ Lancaster Sch Comp & Commun Lancaster LA1 4YW England Univ Leeds Sch Comp Leeds LS2 9JT W Yorkshire England

To accelerate the training of Deep Learning (DL) models, clusters of machines equipped with hardware accelerators such as GPUs are leveraged to reduce execution time. State-of-the-art resource managers are needed to increase GPU utilization and maximize throughput. While co-locating DL jobs on the same GPU has been shown to be effective, this can incur interference causing slowdown. In this article we propose Horus: an interference-aware and prediction-based resource manager for DL systems. Horus proactively predicts GPU utilization of heterogeneous DL jobs extrapolated from the DL model's computation graph features, removing the need for online profiling and isolated reserved GPUs. Through micro-benchmarks and job co-location combinations across heterogeneous GPU hardware, we identify GPU utilization as a general proxy metric to determine good placement decisions, in contrast to current approaches which reserve isolated GPUs to perform online profiling and directly measure GPU utilization for each unique submitted job. Our approach promotes high resource utilization and makespan reduction;via real-world experimentation and large-scale trace driven simulation, we demonstrate that Horus outperforms other DL resource managers by up to 61.5 percent for GPU resource utilization, 23.7-30.7 percent for makespan reduction and 68.3 percent in job wait time reduction.

关键词： Graphics processing units Interference Kernel Predictive models Computational modeling Production Load modeling distributed systems deep learning interference GPU utilization cloud computing workload prediction

来源：评论

学校读者我要写书评

暂无评论

2023 8th IEEE workshop on the Electronic Grid, eGRID 2023

2023 8th IEEE Workshop on the Electronic Grid, eGRID 2023

引用

8th IEEE workshop on the Electronic Grid, eGRID 2023

ISBN: (纸本)9798350327007

The proceedings contain 70 papers. The topics discussed include: defining and constraining the electrical cardinality of multiport converter mission profiles;a scalable transmission and distribution co-simulation platform for IBR-heavy power systems;low-capacitance cascaded H-bridge STATCOM with enhanced control scheme under unbalanced grid;experimental validation of demand side response rates for frequency control;improving power system resilience based on grid-forming converter control and real-time monitoring;application of model-free control to reduce the total harmonic distortion of inverters;composing power flow patterns through coordinated dispatch of energy packets;improved short circuit behavior by distributed capacitors in dc microgrids;and fault-tolerant scheme for modular multilevel converter based on exchanging SMS between arms.

关键词：

来源：评论

学校读者我要写书评

暂无评论

timestamped State Sharing for Stream Analytics

引用

IEEE TRANSACTIONS ON parallel AND distributed systems 2021年第11期32卷 2691-2704页

作者： Zhao, Yunjian Liu, Zhi Wu, Yidi Jiang, Guanxian Cheng, James Liu, Kunlong Yan, Xiao Chinese Univ Hong Kong Dept Comp Sci & Engn Hong Kong Peoples R China

State access in existing distributed stream processing systems is restricted locally within each operator. However, in advanced stream analytics such as online learning and dynamic graph analytics, enabling state sharing across different operators makes application development easier and stream processing more efficient. In addition, when stream records are timestamped, proper time semantics should be defined for both state updates and fetches. We propose a new state abstraction to address the limitations of existing systems and develop a distributed stream processing system, Nova, with native support for timestamped state sharing. We validate the expressiveness and efficiency of Nova with extensive experiments.

关键词： Semantics Pattern matching Throughput Sparks real-time systems Industries Storms State sharing distributed stream processing online learning dynamic graph analytics

来源：评论

学校读者我要写书评

暂无评论

13th Mining Humanistic Data workshop, MHDW 2024, 9th workshop on 5G-Putting Intelligence to the Network Edge, 5G-PINE 2024 and 1st workshop on AI in Applications for Achieving the Green Deal Targets, AI4GD 2024 held as parallel events of the IFIP WG 12.5 International workshops on Artificial Intelligence Applications and Innovations, AIAI 2024

13th Mining Humanistic Data Workshop, MHDW 2024, 9th Worksho...

引用

13th Mining Humanistic Data workshop, MHDW 2024, 9th workshop on 5G-Putting Intelligence to the Network Edge, 5G-PINE 2024 and 1st workshop on AI in Applications for Achieving the Green Deal Targets, AI4GD 2024 held as parallel events of the IFIP WG 12.5 International workshops on Artificial Intelligence Applications and Innovations, AIAI 2024

ISBN: (纸本)9783031632266

The proceedings contain 34 papers. The special focus in this conference is on Mining Humanistic Data. The topics include: Boosting Data Monetisation with DATAMITE;Boosting Digitalization Across European Regions: The AMBITIOUS Approach;Data Monetization Opportunities and Challenges: The European Landscape by DATAMITE;Innovative Digital Forensic and Investigation Tools for Law Enforcement: The EMPOWER & TRACY Approach;market Analysis of a Data Platform in the European Data Ecosystem;OCTAPUS: Empowering Next Generation Central Offices with Cross-Layer Network Intelligence Towards a Reconfigurable Data Plane;On 6G-Enabled SDN-Based Mobile Network User Plane with DRL-Based Traffic Engineering;on the Efficient Architecture for 6G System;Promoting Deployment of Innovative Use Cases in Market Verticals for the Support of 6G Evolution: The 6G-PATH Context;scalable Data Profiling for Quality Analytics Extraction;seamless Integration of Efficient 6G Wireless Technologies for Communication and Sensing Enabling Ecosystems;testing Plan Description and Field Measurements for real-time Wide Area Monitoring of Interconnected Power systems in the Smart5Grid Project;comparative Analysis of time Series and Machine Learning Models for Air Quality Prediction Utilizing IoT Data;Efficient Energy Disaggregation Using DBSCAN: A Novel Approach for Enhanced Energy Management;integrating Machine Learning and Biological Context for Single-Cell Gene Regulatory Network Inference;museum Education: Integration of Cultural Heritage and Educational Metadata Schemas;net Zero Strategies: Empowering Climate Change Solutions Through Advanced Analytics and time Series;New Perspectives in e-Learning: EEG-Based Modelling of Human Cognition Individual Differences;optimizing Vehicle-to-Vehicle Energy Sharing with Predictive Modeling;predicting Song Popularity Through Machine Learning and Sentiment Analysis on Social Networks;transformer-Based Anomaly Detection in Energy Consumption Data;Advancing Agricultural

关键词：

来源：评论

学校读者我要写书评

暂无评论

Study on distributed Consistent Cooperative Control of Multi-ART in Automated Container Terminals

引用

IEEE ACCESS 2022年 10卷 122965-122980页

作者： Zhang, Yu Yang, Caiyun Tang, Kexin Dai, Jinshan Wuhan Univ Technol Sch Transportat & Logist Engn Wuhan 430063 Hubei Peoples R China Inland Port & Shipping Ind Res Co Ltd Shaoguan 512100 Guangdong Peoples R China Minist Educ Engn Res Ctr Port Logist Technol & Equipment Wuhan 430063 Hubei Peoples R China

In order to address the congestion problem of vehicles in quay of automated container terminal with parallel layout and side-loading operations, the study regards the terminal operation system as a multi-agent system (MAS). Then, A dynamic cooperative speed regulation strategy for a group intelligence-oriented multi-ART (Artificial Intelligence Robot of Transportation) is proposed to realize the smart and distributed traffic control of container terminals. Via the real-time data interchange and message passing with ART group, the individual ART implements the collaborative speed regulation with the combination of the dynamic speed regulation strategy. Through the above-mentioned method, the sequence of ARTs arriving at the quay can be adjusted, which can decrease the waiting time of ARTs at the quay. Considering the consistency problem of individual and group agent among the dynamic speed regulation of ARTs, a novel distributed consensus protocol is established to assure that the individual status of ART will tend to converge basically during the collaborative speed regulation process, and the convergence of system can be guaranteed. Based on the case of automated container terminal of Tianjin Port, the paper establishes a multi-agent based simulation model, and simulates the decision-making process, and validates the effectiveness of mentioned model and strategy. The results demonstrate that the strategy can decrease the waiting time of ARTs under uncertainty environment of terminal and improve the operation efficiency of system.

关键词： Agent-based simulation automated container terminals dynamic speed regulation multi-agent systems transport operations

来源：评论

学校读者我要写书评

暂无评论

neuroAIx: FPGA Cluster for Reproducible and Accelerated Neuroscience Simulations of SNNs 9

neuroAIx: FPGA Cluster for Reproducible and Accelerated Neur...

引用

9th IEEE Nordic Circuits and systems Conference (NorCAS)

作者： Kauth, Kevin Stadtmann, Tim Sobhani, Vida Gemmeke, Tobias Rhein Westfal TH Aachen Chair Integrated Digital Syst & Circuit Design ID Aachen Germany

ISBN: (纸本)9798350337570

The complex mechanisms and impressive learning capabilities of the human brain inspire the search for ever more efficient neural network models. Investigating the specific nature of biological neural networks in accelerated simulations with hundreds of millions of synapses, however, pushes even the most sophisticated systems to their limits. At the same time, the inherent chaotic nature of spiking networks and their typically distributed simulation complicates reproducibility, which is a major concern in both research and engineering. With HPC systems incrementally surpassing biological real-time in the simulation of neuroscience networks of significant size and complexity, novel computing architectures are required to reach disruptive speed-ups while assuring determinism and providing flexibility for continuous adaptations to new insights from neuroscience. To this end, we developed the neuroAIx FPGA cluster, based on off-the-shelf AMD/Xilinx FPGAs. It integrates a proprietary ultra-low-latency communication architecture and a custom short-range synchronization algorithm that brings worst-case source-to-destination spike latency down to 2 microseconds. Together with a latency-hiding memory-prefetch procedure, it enables deterministic simulations of the widely-used cortical microcircuit model, with nearly 80k neurons and 300M synapses, with an acceleration factor of 20x compared to biological real-time. This surpasses all existing systems, outperforming non-FPGA platforms by 10x, while providing the flexibility needed to explore neuro-inspired algorithms in the future.

关键词： neuromorphic computing computational neuroscience FPGA cluster parallel computing cortical microcircuit accelerated simulation spiking neural networks (SNN) HPC

来源：评论

学校读者我要写书评

暂无评论

Detecting Performance Variance for parallel Applications Without Source Code

引用

IEEE TRANSACTIONS ON parallel AND distributed systems 2022年第12期33卷 4239-4255页

作者： Zhai, Jidong Zheng, Liyan Zhang, Feng Tang, Xiongchao Wang, Haojie Yu, Teng Jin, Yuyang Song, Shuaiwen Leon Chen, Wenguang Tsinghua Univ Dept Comp Sci & Technol Beijing 100084 Peoples R China Renmin Univ China Sch Informat Key Lab Data Engn & Knowledge Engn Beijing 100872 Peoples R China Univ Sydney Sch Comp Sci Sydney NSW 2006 Australia

For parallel applications, performance variance is a critical issue that can degrade performance and make applications' behavior difficult to explain. Therefore, users and application developers should be able to detect and diagnose performance variance. Previous detection methods either introduce too much overhead and slow down applications, or rely on nontrivial source code analysis, which is impractical for production-run parallel systems. In this article, we propose Vapro, a framework for detecting and diagnosing performance variance in production-run parallel systems. Our method is based on an observation that most parallel programs contain code snippets that are executed repeatedly with a fixed workload and can be utilized to detect performance variance. We present State Transition Graph (STG) to track program execution and then do light-weight workload analysis on STG to locate performance variance. Vapro is able to successfully identify these snippets at runtime even without program source code. To diagnose the discovered variation, Vapro uses a progressive diagnosis method based on a hybrid model combining variance breakdown and statistical analysis. According to evaluating results, Vapro's performance overhead is only 1.38% on average. Vapro can identify performance variance in real applications caused by hardware issues, such as memory and IO. The standard deviation of the execution time is decreased by up to 73.5% when the identified variance is fixed. Vapro achieves 30.0% larger detection coverage than the state-of-the-art variance detection approach based on source code analysis.

关键词： Codes Runtime Hardware Performance evaluation Libraries Benchmark testing Electric breakdown Performance variance anomaly detection system noise parallel computing

来源：评论

学校读者我要写书评

暂无评论

Towards Ubiquitous Intelligent Computing: Heterogeneous distributed Deep Neural Networks

引用

IEEE TRANSACTIONS ON BIG DATA 2022年第3期8卷 644-657页

作者： Zhang, Zongpu Song, Tao Lin, Liwei Hua, Yang He, Xufeng Xue, Zhengui Ma, Ruhui Guan, Haibing Shanghai Jiao Tong Univ Sch Elect Informat & Elect Engn Shanghai Key Lab Scalable Comp & Syst Shanghai 200240 Peoples R China United Kingdom EEECS ECIT Belfast BT3 9DT Antrim North Ireland

For the pursuit of ubiquitous computing, distributed computing systems containing the cloud, edge devices, and Internet-of-Things devices are highly demanded. However, existing distributed frameworks do not tailor for the fast development of Deep Neural Network (DNN), which is the key technique behind many intelligent applications nowadays. Based on prior exploration on distributed deep neural networks (DDNN), we propose Heterogeneous distributed Deep Neural Network (HDDNN) over the distributed hierarchy, targeting at ubiquitous intelligent computing. While being able to support basic functionalities of DNNs, our framework is optimized for various types of heterogeneity, including heterogeneous computing nodes, heterogeneous neural networks, and heterogeneous system tasks. Besides, our framework features parallel computing, privacy protection and robustness, with other consideration for the combination of heterogeneous distributed system and DNN. Extensive experiments demonstrate that our framework is capable of utilizing hierarchical distributed system better for DNN and tailoring DNN for real-world distributed system properly, which is with low response time, high performance, and better user experience.

关键词： Neural networks Cloud computing Ubiquitous computing Task analysis Computational modeling Big Data heterogeneous distributed deep neural network HDDNN deep neural network DNN Internet of Things edge computing cloud computing

来源：评论

学校读者我要写书评

暂无评论

Endpoint Architecture for distributed real-time Applications Based on TSN

Endpoint Architecture for Distributed Real-Time Applications...

引用

IEEE International workshop on Factory Communication systems (WFCS)

作者： Stefan Oechsle Florian Frick Moritz Walker Armin Lechler Alexander Verl Institute for Control Engineering of Machine Tools and Manufacturing Units (ISW) University of Stuttgart Stuttgart Germany

ISBN: (数字)9798350319347

ISBN: (纸本)9798350319354

distributed real-time applications perform deterministic tasks across multiple connected compute platforms, referred to as endpoints. Deterministic behavior of the communication and the compute are prerequisites, but coordinating and managing both in a holistic way is a significant challenge. Converged real-time networks based on time-Sensitive Networking (TSN) are considered the key enabler across industries for the required connectivity. To date, the potential of TSN cannot be exploited due to many sub-standards still being under development, gaps in the management and configuration functions, lack of best practices and unresolved integration with higher application layers and domain specific engineering. Previous work, focusing on a modular open source framework for the engineering and management of distributed real-time applications, identified a unified approach for TSN-based endpoints as a key gap. This paper presents a novel endpoint architecture to operate, engineer and manage distributed real-time applications by seamlessly integrating real-time networking and compute. The architecture provides interfaces to a central engineering, allows to integrate centralized as well as distributed network management and offers a unified data management. An abstraction layer for applications allows the efficient use and complements existing open source solutions.

关键词： Industries Communication systems Focusing distributed databases Computer architecture real-time systems Production facilities

来源：评论

学校读者我要写书评

暂无评论

An Intelligent parallel distributed Streaming Framework for near real-time Science Sensors and High-Resolution Medical Images 21

An Intelligent Parallel Distributed Streaming Framework for ...

引用

50th International Conference on parallel Processing (ICPP)

作者： Shivadekar, Samit Mangalagiri, Jayalakshmi Phuong Nguyen Chapman, David Halem, Milton Gite, Rahul Univ Maryland Baltimore Cty Baltimore MD 21228 USA OpenKneck Inc Halethorpe MD USA

ISBN: (纸本)9781450384414

Our goals are to address challenges such as latency, scalability, throughput and heterogeneous data sources of streaming analytics and deep learning pipelines in science sensors and medical imaging applications. We present a prototype Intelligent parallel distributed Streaming Framework (IPDSF) that is capable of distributed streaming processing as well as performing distributed deep training in batch mode. IPDSF is designed to run streaming Artificial Intelligent (AI) analytic tasks using data parallelism including partitions of multiple streams of short time sensing data and high-resolution 3D medical images, and fine grain tasks distribution. We will show the implementation of IPDSF for two real world applications, (i) an Air Quality Index based on near real time streaming of aerosol Lidar backscatter and (ii) data generation of Covid-19 Computing Tomography (CT) scans using deep learning. We evaluate the latency, throughput, scalability, and quantitative evaluation of training and prediction compared against a baseline single instance. As the results, IPDSF scales to process thousands of streaming science sensors in parallel for Air Quality Index application. IPDSF uses novel 3D conditional Generative Adversarial Network (cGAN) training using parallel distributed Graphic Processing Units (GPU) nodes to generate realistic 3D high resolution Computed Tomography scans of Covid-19 patient lungs. We will show that IPDSF can reduce cGAN training time linearly with the number of GPUs.

关键词： distributed streaming Data parallelism Deep Learning Lidar backscatter Computed Tomography cGAN

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：