检索结果-内蒙古大学图书馆

30th international conference on Architectural Support for Programming Languages and Operating Systems-ASPLOS

作者： Tan, Yifan Tan, Cheng Mi, Zeyu Chen, Haibo Shanghai Jiao Tong Univ SEIEE Inst Parallel & Distributed Syst Shanghai Peoples R China Northeastern Univ Boston MA USA

ISBN: (纸本)9798400706981

Confidential computing on GPUs, like NVIDIA H100, mitigates the security risks of outsourced Large Language Models (LLMs) by implementing strong isolation and data encryption. Nonetheless, this encryption incurs a significant performance overhead, reaching up to 52.8% and 88.2% throughput drop when serving OPT-30B and OPT-66B, respectively. To address this challenge, we introduce PipeLLM, a user-transparent runtime system. PipeLLM removes the overhead by overlapping the encryption and GPU computation through pipelining-an idea inspired by the CPU instruction pipelining-thereby effectively concealing the latency increase caused by encryption. The primary technical challenge is that, unlike CPUs, the encryption module lacks prior knowledge of the specific data needing encryption until it is requested by the GPUs. To this end, we propose speculative pipelined encryption to predict the data requiring encryption by analyzing the serving patterns of LLMs. Further, we have developed an efficient, low-cost pipeline relinquishing approach for instances of incorrect predictions. Our experiments show that compared with vanilla systems without confidential computing (e.g., vLLM, PEFT, and FlexGen), PipeLLM incurs modest overhead (<19.6% in throughput) across various LLM sizes, from 13B to 175B. PipeLLM's source code is available at https://***/SJTU-IPADS/PipeLLM.

关键词： Nvidia Confidential computing Large Language Model Confidential Virtual Machine

来源：评论

学校读者我要写书评

暂无评论

Research on key technologies of cloud-edge collaboration in the power IoT based on computing power perception

Research on key technologies of cloud-edge collaboration in ...

引用

2024 international conference on Computer Graphics, Artificial Intelligence, and Data Processing, ICCAID 2024

作者： Chen, Wu State Grid Info-Telecom Great Power Science and Technology Co. Ltd. Fuzhou350101 China

ISBN: (纸本)9781510689275

In the construction process of new power systems, the concept of power cluster control has emerged with the integration of large-scale distributed new energy, and regional autonomy is an effective means of distributed new energy consumption. At present, the electric power smart IoT focuses on the collection and distribution of environmental, state, and low-voltage electrical quantities in various scenarios of transmission and distribution in quasi real-time. However, the edge IoT technology that supports distributed new energy consumption in substations and flexible coordinated control between substations is not yet perfect. The traditional processing method relying on centralized cloud platforms has poor real-time performance. The existing electric power smart IoT system has limited edge computing resources, which are difficult to meet the rapid response needs of single substation autonomy and multi substation interaction. There is an urgent need for resource collaboration on the edge and platform sides, and research on key technologies for cloud edge collaboration of electric power IoT based on computing power perception is needed to expand and improve the existing electric power smart IoT system. © 2025 SPIE.

关键词： Edge computing

来源：评论

学校读者我要写书评

暂无评论

distributed Primal-Dual Interior Point Framework for Analyzing Infeasible Combined Transmission and Distribution grid Networks 58

Distributed Primal-Dual Interior Point Framework for Analyzi...

引用

58th Hawaii international conference on System Sciences, HICSS 2025

作者： Ali, Muhammad Hamza Pandey, Amritanshu University of Vermont United States

ISBN: (纸本)9780998133188

The proliferation of distributed energy resources has heightened the interactions between transmission and distribution (T&D) systems, necessitating novel analyses for the reliable operation and planning of interconnected T&D networks. A critical gap is an analysis approach that identifies and localizes the weak spots in the combined T&D networks, providing valuable information to system planners and operators. The research goal is to efficiently model and simulate infeasible (i.e. unsolvable in general settings) combined positive sequence transmission and three-phase distribution networks with a unified solution algorithm. We model the combined T&D network with the equivalent circuit formulation. To solve the overall T&D network, we build a Gauss-Jacobi-Newton (GJN) based distributed primal dual interior point optimization algorithm capable of isolating weak nodes. We validate the approach on large combined T&D networks with 70k+ T and 15k+ D nodes and demonstrate performance improvement over the alternating direction method of multipliers (ADMM) method. © 2025 IEEE Computer Society. All rights reserved.

关键词： grid computing

来源：评论

学校读者我要写书评

暂无评论

Research on grid-Connected Model of distributed Generation Based on Colored Petri Net 1

引用

5th international conference on Edge computing (EDGE) Held as Part of the Services conference Federation (SCF)

作者： Chen, Bin Han, Qiang Xu, Huan Guo, Yuxiang Li, Jie Song, Tairan China Southern Power Grid Co Ltd Big Data Digitalizat Dept Guangzhou 510663 Guangdong Peoples R China Macau Univ Sci & Technol Macau Peoples R China

ISBN: (数字)9783030965044

ISBN: (纸本)9783030965044;9783030965037

With the strategy of "Emission Peak" and "Carbon Neutrality" proposed the transformation and reform of the energy system are imminent Building a new power system mainly composed of renewable power is one of the most important measures to achieve the dual goals on carbon. The stability and reliability of the new power system have become the focus of attention. Therefore, it is necessary to minimize the impact of distributed generation integration on the power system and find out the optimal configuration model. By studying the working mechanism of distributed generation such as wind power generation, photovoltaic power generation and energy storage power supply, this paper studies and summarizes the working strategy and operation mode during distributed power grid-connected and islanded operation, and further designs the distributed generation integration model based on Colored Petri Net for optimization. By deducing the configuration optimization of the grid connection model from the grid connection access point, network loss and economy, the analysis logic of the optimal grid connection access point, optimal grid connection capacity and optimal configuration is finally output. The method deduction shows that the grid connection model provides a new grid connection optimization method for distributed generation grid connection, which makes up for the relative instability of traditional grid connection in power system.

关键词： New power system distributed generation grid-connected model Colored Petri Net

来源：评论

学校读者我要写书评

暂无评论

Research on Sustainable Management of grid Companies Based on Artificial Intelligence 2

Research on Sustainable Management of Grid Companies Based o...

引用

2nd IEEE international conference on distributed computing and Electrical Circuits and Electronics, ICDCECE 2023

作者： Liu, Xiaoyu Yue, Guanghui Shan, Bingzhe State Grid Hunan Electric Power Company Hunan Changsha China

ISBN: (纸本)9798350347456

With the advancement of power systems, grid operations monitoring has become more complex. To ensure the sustainable operation of the grid, it is important to quickly and dynamically extract a small amount of data from a large amount of real-time data, closely related to grid security, and accurately identify system security risks in a timely manner. In this regard, this paper proposes an artificially intelligent sustainable management system for the grid, which has achieved remarkable results. The main contribution of this paper is the development of a risk identification model for microgrids using artificial intelligence and the DEMATEL method. Firstly, the paper identifies the risk factors involved in the microgrid from four aspects, including the generation side, distribution side, demand side, and human factors, through a comprehensive literature review, expert survey, and brainstorming. Secondly, an artificial intelligence and DEMATEL-based microgrid risk factor identification method is employed to clarify the importance and perturbation relationship of each risk factor of the microgrid. Finally, the paper classifies all factors into two categories, cause factors, and effect factors, and ranks the importance of each factor. To further demonstrate the effectiveness of the proposed model, a wind power prediction algorithm based on data mining technology and an improved SVM algorithm, and a PV power prediction algorithm based on a deep neural network are established. After comparing and analyzing the performance of the constructed algorithms with other algorithms, the DBN prediction model is proposed, which has an absolute error probability of 62.9% within 1%, surpassing the other algorithms and meeting the engineering needs. Moreover, the paper proposes risk control measures on the power generation side, which can significantly reduce the risks involved in power generation. In summary, the paper proposes an artificially intelligent sustainable management system

关键词： Smart power grids

来源：评论

学校读者我要写书评

暂无评论

A Genetic Algorithm based parallel Task Planning Method for Cooperative MRS 26

A Genetic Algorithm based Parallel Task Planning Method for ...

引用

26th ACIS international Winter conference on Software Engineering, Artificial Intelligence, Networking and parallel/distributed computing, SNPD-Winter 2023

作者： Yu, Li Tianyang, Wang Zhongqiu, Zhang Xiaokun, Ren Northwestern Polytechnical University Xi'an China School of Computer Northwestern Polytechnical University Xi'an China Center of Information Northwestern Polytechnical University Xi'an China

ISBN: (纸本)9798350345865

With in the trend of intelligence, task cooperation has been emerging as one main feature for Multiple Robot Systems (MRS) and applications, typically Multi-UAV (MUAV) system. For such swarm cooperative systems, how to plan and map their tasks to proper members within systems optimally and fast has been one vital issue that affects the efficiency of cooperation. So, in this article, a two-level task model is firstly established, which is the foundation of automatic task parsing and planning. Further, a Genetic Algorithm (GA) based parallel Task Planning algorithm PTP-GA is proposed and designed, the goal of which is to improve the performance via cluster related tasks together and separating the dividing the solution space. Particularly, several constrained operators are designed to warrantee both the randomness and elitist individuals or segments. Finally, this proposed method is verified within a simulated scenario, and it shows that the performance of this parallel algorithm is indeed improved. © 2023 IEEE.

关键词： Genetic algorithms

来源：评论

学校读者我要写书评

暂无评论

HPC Performance on AWS: EFA Integrated Intel Instances for MPI Applications

HPC Performance on AWS: EFA Integrated Intel Instances for M...

引用

international conference on parallel distributed and grid computing (PDGC)

作者： Mandeep Kumar Gagandeep Kaur Prashant Singh Rana University Institute of Computing Chandigarh University Mohali India Locuz Enterprise Solutions Private Limited Delhi India Computer Science and Engineering Department Thapar Institute of Engineering and Technology Patiala India

ISBN: (数字)9798331521349

ISBN: (纸本)9798331521356

High Performance computing (HPC) has become an indispensable tool in various scientific and engineering domains, demanding efficient infrastructure to support computationally intensive tasks. Cloud computing platforms, such as Amazon Web Services (AWS), have emerged as a viable solution for provisioning HPC resources. In this work, we explore the performance of HPC workloads running on AWS, utilizing Elastic Fabric Adapter (EFA) interconnect technology and Intel Xeon Scalable Processors. The investigation focuses on Message Passing Interface (MPI) applications, evaluating the system's efficiency, scalability, and the benefits of EFA integration. We present experimental results and insights into optimizing HPC workloads in the AWS environment, contributing to the ongoing discussion on achieving high-performance computational capabilities in the cloud.

关键词： Cloud computing Program processors Web services Software packages Scalability Message passing Software algorithms Throughput Resource management Optimization

来源：评论

学校读者我要写书评

暂无评论

MANTRA: an Edge-computing Framework based on Multi-Armed Bandit for Latency- and Energy-aware Job Offloading in Vehicular Networks 9

MANTRA: an Edge-Computing Framework based on Multi-Armed Ban...

引用

9th IEEE international conference on Network Softwarization (IEEE NetSoft) - Boosting Future Networks through Advanced Softwarization

作者： Busacca, Fabio Palazzo, Sergio Raftopoulos, Raoul Schembra, Giovanni Univ Catania CNIT Res Unit Catania Italy

ISBN: (纸本)9798350399806

The optimization of job offloading procedures in modern vehicular networks is a problem of utmost importance. In this regard, this paper proposes MANTRA, a distributed framework based on multi-player multi-armed bandit (MPMAB) algorithms for latency- and energy-aware job offloading in vehicular networks. The main goal of MANTRA is to support procedures of job offloading in green vehicular networks to achieve a target tradeoff between energy consumption and job processing latency. In particular, MANTRA is intended to run on so-called MEC-in-a-box (M-Box) devices, portable batterypowered Road Side Units (RSUs) specifically designed to work without mobile connectivity and of a fixed power grid. To demonstrate MANTRA effectiveness, we model the vehicular network using the queueing theory for M/M/m/K systems. We run an extensive evaluation campaign and compare MANTRA with several baselines, including a centralized, oracle-based approach. In such a way, we demonstrate how MANTRA outperforms the baselines and quickly converges to the performance of the centralized approach in a fully-distributed way in terms of job processing latency and network outage probability.

关键词： Edge computing Vehicular Networks Multiarmed Bandit Job Offloading

来源：评论

学校读者我要写书评

暂无评论

Discrete Geometric Coded Data Layout for Large-scale Object Storage Systems 21

Discrete Geometric Coded Data Layout for Large-scale Object ...

引用

21st IEEE international Symposium on parallel and distributed Processing with Applications, 13th IEEE international conference on Big Data and Cloud computing, 16th IEEE international conference on Social computing and Networking and 13th international conference on Sustainable computing and Communications, ISPA/BDCloud/SocialCom/SustainCom 2023

作者： Tian, Yi Xu, Guangping Yang, Hongzhang Ni, Yue Cao, JiaXin Yang, Lei Tianjin University of Technology School of Computer Science and Engineering Tianjin China Novel Software Technology Tianjin Key Laboratory of Intelligence Computing Tianjin China Key Laboratory of Computer Vision and System Ministry of Education China Information Technology Co. Ltd. China

ISBN: (纸本)9798350329223

Regenerating codes are new network codes proposed to reduce the data required for fault repair, which can improve the recovery efficiency of faulty nodes in data storage systems. However, unlike Reed-Solomon code, which repairs at the granularity of bytes, regenerating codes require data stored in large chunks but leads to severe read amplification, which reads out excess data when degraded read objects and increases degraded read time. That reflects a mutual constraint between improving recovery efficiency and degraded read performance, as manifested in the amplification of data reads, a vital issue considered in this *** solve this problem, we propose a new type of data layout - discrete geometric, which splits the object into a series of geometric sequences of data blocks. They are placed discretely into corresponding containers on the disks at different nodes, with containers of the same size made into a strip for encoding. The discrete characteristic ensures lower repair costs for degraded reads. The geometric characteristic ensures the repair performance of regenerating codes by large blocks, and read amplification can be mitigated through small blocks. To reduce IOPS for discrete geometric, we propose Discrete Geometric-Locally Regenerating Codes (DG-LRCs), guaranteeing lower degraded read latency while improving recovery *** results show that the degraded read time of DG-LRCs compared to regenerating codes combined with geometric partitioning is 22.56% lower at 2Gbpsand 60.56% lower at 4Gbps, and the recovery performance is 7.04 times better than that of RS code. © 2023 IEEE.

关键词： Efficiency

来源：评论

学校读者我要写书评

暂无评论

Optimal locating method of edge computing device in cyber physical distribution system

引用

ENERGY REPORTS 2022年 8卷 684-694页

作者： Zhang, Jian Yang, Ting Ji, Hao Li, Wei Tianjin Univ Key Lab Smart Grid Minist Educ Tianjin 300072 Peoples R China Tianjin Elect Power Co State Grid Tianjin 300010 Peoples R China Univ Sydney Sch Comp Sci Sydney NSW 2006 Australia

With the development of interconnection of massive terminals, high maintenance of equipment data and complexity of network structure in distribution system, the traditional information processing mode based on single cloud master station has some disadvantages, such as high data transmission delay, high congestion, information far away from equipment and slow queuing processing. Edge computing can make local fast analysis and decision for data in the near terminal device side, which can effectively make up for the deficiency of cloud computing. As the cornerstone of edge computing, edge control processing node is the first premise to realize fast response control of cyber physical distribution system (CPDS). In this paper, a new method of edge computing node deployment is proposed for the distributed edge computing requirements of cyber physical distribution system, considering the cyber stability and power stability. The community theory based on distribution network power flow and line impedance is used to divide the edge computing autonomous region, and the calculation model of cyber stability and power stability of CPDS network is established, then the optimal deployment of CPDS edge computing nodes is solved by the hybrid entropy calculation of cyber and power. Finally, an improved standard example is used for simulation analysis. The simulation results show that the ECN deployment scheme proposed in this paper can effectively reduce network congestion, improve network response speed and realize optimal control of distribution network. (C) 2022 The Authors. Published by Elsevier Ltd.

关键词： Integrated solar combined cycle Inertial power compensation Distribution network Cyber physical system Edge computing Stability The hybrid entropy of cyber and power

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：