Load unbalancing problem has a significant impact on the parallel efficiency of fluid-structure interaction simulation in cavitating flow. When the total parallelism is determined, the speedup will be seriously affect...
详细信息
This article proposes a method for detecting the distance between power transformer bushings based on binocular stereo vision, which utilizes the Grabcut image segmentation algorithm to achieve intelligent segmentatio...
详细信息
The fine-grained calculation method of accepting distributed generation capacity in distribution network for virtual power plant is proposed in this paper. Firstly, a distributed generation capacity model including ph...
详细信息
As nonrenewable fuels deplete;most of the businesses are transitioning to renewable energy. Solar energy is a abundant in nature and is a reliable source of renewable energy. In today's photovoltaic systems, multi...
详细信息
Islanding detection becomes a necessity when the DERs units are required to continue their generation even after the islanding of the μG from the grid. This paper investigates the effectiveness of the active islandin...
详细信息
A framework to support optimised application placement across the cloud-edge continuum is described, making use of the Optimized-Greedy Nominator Heuristic (EO-GNH). The framework can be employed across a range of dif...
详细信息
ISBN:
(纸本)9783031506833;9783031506840
A framework to support optimised application placement across the cloud-edge continuum is described, making use of the Optimized-Greedy Nominator Heuristic (EO-GNH). The framework can be employed across a range of different Internet of Things (IoT) applications, such as smart agriculture and healthcare. The framework uses asynchronous MapReduce and parallel meta-heuristics to support the management of IoT applications, focusing on metrics such as execution performance, resource utilization and system resilience. We evaluate EOGNH using service quality achieved through real-time resource management, across multiple application domains. Performance analysis and optimisation of EO-GNH has also been carried out to demonstrate how it can be configured for use across different IoT usage contexts.
In this project, a new type of electric ground rod suitable for temporary ground wire operation is developed. It can be combined with the standard ground wire operating lever used in substation to complete the reliabl...
详细信息
A method for generating and reducing distributed power generation output scenarios based on improved clustering analysis is proposed to address the issues of low accuracy and susceptibility to local optima in typical ...
详细信息
Modern advancements in large-scale machine learning would be impossible without the paradigm of data-paralleldistributedcomputing. Since distributedcomputing with large-scale models imparts excessive pressure on co...
详细信息
Modern advancements in large-scale machine learning would be impossible without the paradigm of data-paralleldistributedcomputing. Since distributedcomputing with large-scale models imparts excessive pressure on communication channels, significant recent research has been directed toward co-designing communication compression strategies and training algorithms with the goal of reducing communication costs. While pure data parallelism allows better data scaling, it suffers from poor model scaling properties. Indeed, compute nodes are severely limited by memory constraints, preventing further increases in model size. For this reason, the latest achievements in training giant neural network models also rely on some form of model parallelism. In this work, we take a closer theoretical look at Independent Subnetwork Training (IST), which is a recently proposed and highly effective technique for solving the aforementioned problems. We identify fundamental differences between IST and alternative approaches, such as distributed methods with compressed communication, and provide a precise analysis of its optimization performance on a quadratic model. Copyright 2024 by the author(s)
The surge in demand for computing resources in data centers coupled with the rise of environmental concerns has motivated cloud providers to reduce carbon emission due to computational energy consumption. An opportuni...
详细信息
ISBN:
(纸本)9798331531317;9798331531300
The surge in demand for computing resources in data centers coupled with the rise of environmental concerns has motivated cloud providers to reduce carbon emission due to computational energy consumption. An opportunity lies in the fluctuating availability of renewable energy over time and the variability of power sources over grid regions, leading to variations in space and time in carbon intensity. Exploiting such variations, this paper introduces Caspian, a carbon-aware workload scheduler in multi-cluster Kubernetes environments, which aims at reducing the Carbon Footprint (CFP) due to executing workloads, while satisfying Quality of Service (QoS) requirements. Caspian cooperates with a multi-cluster management platform to apply scheduling and placement decisions over distributed clusters. We present efficient optimization algorithms to achieve these goals. Further, we describe an implementation of Caspian, integrated with Multi Cluster App Dispatcher (MCAD), a multi-cluster management platform which handles queuing and dispatching of workloads over multiple clusters. Our experimental results show that Caspian effectively reduces CFP with reasonable QoS, compared to a baseline scheduler which only satisfies the QoS of workloads. Specifically, Caspian reduces CFP by about 33%, with about 98% of workloads completing at an average fraction of 0.6 of their deadline.
暂无评论