Based on the current Service-Oriented Architecture （SOA）, this paper proposes a new collaborative software design methodology in an SOA Environment： the Global Software Enterprise （GSE）. The current SOA is produc...Based on the current Service-Oriented Architecture （SOA）, this paper proposes a new collaborative software design methodology in an SOA Environment： the Global Software Enterprise （GSE）. The current SOA is producer-centric, in which the service providers publish services that they produce and let the consumers to search avail- able services to compose their applications. GSE is build on top of Consumer-Centric SOA （CCSOA）, in which the application builders publish their application requirements for the service providers to follow when producing or customizing services to support the application. This new methodology reduces the workload and improves the application description capability of the service consumers. It also extends the capacity of design and code sharing, and thus further improves the software productivity. This paper presents the concepts, architecture, enabling techniques, and illustrative examples of collaborative software design in GSE.展开更多
AI大模型正引领下一个十年的信息与通信技术(information and communications technology,ICT)产业发展热点。智算中心网络是支撑AI大模型分布式训练的通信底座,是决定AI集群效能的关键要素之一。AI大模型的数据量和参数量不断扩张,给...AI大模型正引领下一个十年的信息与通信技术(information and communications technology,ICT)产业发展热点。智算中心网络是支撑AI大模型分布式训练的通信底座,是决定AI集群效能的关键要素之一。AI大模型的数据量和参数量不断扩张,给智算中心网络带来了严峻的挑战,同时给关键网络技术进行代际性创新带来了机遇。在AI大模型训练和推理过程中,提供数据的高性能和高安全传输是AI业务对智算中心网络的两大核心需求。高效的负载均衡、拥塞控制技术和网络安全协议是其中的关键网络技术。为应对大规模AI业务带来的严峻挑战,提出全调度以太网(global scheduled Ethernet,GSE)作为对应的解决方案,并搭建真实的测试环境对GSE和RoCE(remote direct memory access over converged Ethernet)网络进行性能对比测试。测试结果证明,GSE相较RoCE网络显著改善了任务完成时间(job completion time,JCT)。展开更多
The modeling of headway/spacing between two consecutive vehicles in a queue has many applications in traffic flow theory and transport practice. Most known approaches have only studied vehicles on freeways. This paper...The modeling of headway/spacing between two consecutive vehicles in a queue has many applications in traffic flow theory and transport practice. Most known approaches have only studied vehicles on freeways. This paper presents a model for the spacing distribution of queuing vehicles at a signalized junction based on random-matrix theory. The spacing distribution of a Gaussian symplectic ensemble （GSE） fits well with recently measured spacing distribution data. These results are also compared with measured spacing distribution observed for the car parking problem. Vehicle stationary queuing and vehicle parking have different spacing distributions due to different driving patterns.展开更多
文摘Based on the current Service-Oriented Architecture （SOA）, this paper proposes a new collaborative software design methodology in an SOA Environment： the Global Software Enterprise （GSE）. The current SOA is producer-centric, in which the service providers publish services that they produce and let the consumers to search avail- able services to compose their applications. GSE is build on top of Consumer-Centric SOA （CCSOA）, in which the application builders publish their application requirements for the service providers to follow when producing or customizing services to support the application. This new methodology reduces the workload and improves the application description capability of the service consumers. It also extends the capacity of design and code sharing, and thus further improves the software productivity. This paper presents the concepts, architecture, enabling techniques, and illustrative examples of collaborative software design in GSE.
文摘AI大模型正引领下一个十年的信息与通信技术(information and communications technology,ICT)产业发展热点。智算中心网络是支撑AI大模型分布式训练的通信底座,是决定AI集群效能的关键要素之一。AI大模型的数据量和参数量不断扩张,给智算中心网络带来了严峻的挑战,同时给关键网络技术进行代际性创新带来了机遇。在AI大模型训练和推理过程中,提供数据的高性能和高安全传输是AI业务对智算中心网络的两大核心需求。高效的负载均衡、拥塞控制技术和网络安全协议是其中的关键网络技术。为应对大规模AI业务带来的严峻挑战,提出全调度以太网(global scheduled Ethernet,GSE)作为对应的解决方案,并搭建真实的测试环境对GSE和RoCE(remote direct memory access over converged Ethernet)网络进行性能对比测试。测试结果证明,GSE相较RoCE网络显著改善了任务完成时间(job completion time,JCT)。
基金Partly supported by the National Key Basic Research and Development(973) Program of China(No.2006CB705506)the National High-Tech Research and Development(863) Program of China (Nos.2006AA11Z229,2007AA11Z222)the National Natural Science Foundation of China(Nos.60374059,50708055)
文摘The modeling of headway/spacing between two consecutive vehicles in a queue has many applications in traffic flow theory and transport practice. Most known approaches have only studied vehicles on freeways. This paper presents a model for the spacing distribution of queuing vehicles at a signalized junction based on random-matrix theory. The spacing distribution of a Gaussian symplectic ensemble （GSE） fits well with recently measured spacing distribution data. These results are also compared with measured spacing distribution observed for the car parking problem. Vehicle stationary queuing and vehicle parking have different spacing distributions due to different driving patterns.
