平方千米阵列(Square Kilometre Array,SKA)科学数据处理产生的数据超出了所有已存在的分布式处理系统的处理能力,如何实现一个分布式执行框架是当前科学数据处理的一个重要研究内容。Spark是一个非常成熟的商业框架,在互联网中被广泛应...平方千米阵列(Square Kilometre Array,SKA)科学数据处理产生的数据超出了所有已存在的分布式处理系统的处理能力,如何实现一个分布式执行框架是当前科学数据处理的一个重要研究内容。Spark是一个非常成熟的商业框架,在互联网中被广泛应用,根据平方千米阵列项目进展的要求,重点研究了如何将算法参考库(Algorithm Reference Library,ARL)中的部分管线移植到Spark上执行。对部分实现过程进行了分析讨论,给出了相应的任务流程。最终结果表明,移植后代码生成结果符合预期,Spark能够满足部分分布式数据的要求,但迫切需要解决自身存在的一系列问题。展开更多
The key technologies involved in the evolution of the Cloud-based Radio Access Network(C-RAN) are discussed in this paper. Taking the Frameless Network Architecture(FNA) as a starting point, a cell-lessbased network t...The key technologies involved in the evolution of the Cloud-based Radio Access Network(C-RAN) are discussed in this paper. Taking the Frameless Network Architecture(FNA) as a starting point, a cell-lessbased network topology for a multi-tier Heterogeneous Network(Het Net) and ultra-dense network is proposed. The FNA network topology modeling is researched with centralized processing and distributed antenna deployments. The Antenna Element(AE) is released as a new dimensional radio resource that is included in the centralized Radio Resource Management(RRM) processes. This contributes to the on-demand user-centric serving-set associations with cell-edge effect elimination. The Control Plane(CP) and User Plane(UP) separation and adaptation are introduced for energy efficiency improvements. The centralized RRM and different optimization goals are discussed for fully exploring the merits from the centralized computing of C-RAN. Considering the complexity, near-optimal approaches for specific users' Quality-of-Service(Qo S) requirements are addressed. Finally, based on the research highlighted above, the way forward of C-RAN evolution is discussed.展开更多
文摘平方千米阵列(Square Kilometre Array,SKA)科学数据处理产生的数据超出了所有已存在的分布式处理系统的处理能力,如何实现一个分布式执行框架是当前科学数据处理的一个重要研究内容。Spark是一个非常成熟的商业框架,在互联网中被广泛应用,根据平方千米阵列项目进展的要求,重点研究了如何将算法参考库(Algorithm Reference Library,ARL)中的部分管线移植到Spark上执行。对部分实现过程进行了分析讨论,给出了相应的任务流程。最终结果表明,移植后代码生成结果符合预期,Spark能够满足部分分布式数据的要求,但迫切需要解决自身存在的一系列问题。
基金supported by the National High Technology Research and Development Program of China No.2014AA01A701Nature and Science Foundation of China under Grants No.61471068,61421061+2 种基金Beijing Nova Programme No.Z131101000413030International Collaboration Project No.2015DFT10160National Major Project No.2016ZX03001009-003
文摘The key technologies involved in the evolution of the Cloud-based Radio Access Network(C-RAN) are discussed in this paper. Taking the Frameless Network Architecture(FNA) as a starting point, a cell-lessbased network topology for a multi-tier Heterogeneous Network(Het Net) and ultra-dense network is proposed. The FNA network topology modeling is researched with centralized processing and distributed antenna deployments. The Antenna Element(AE) is released as a new dimensional radio resource that is included in the centralized Radio Resource Management(RRM) processes. This contributes to the on-demand user-centric serving-set associations with cell-edge effect elimination. The Control Plane(CP) and User Plane(UP) separation and adaptation are introduced for energy efficiency improvements. The centralized RRM and different optimization goals are discussed for fully exploring the merits from the centralized computing of C-RAN. Considering the complexity, near-optimal approaches for specific users' Quality-of-Service(Qo S) requirements are addressed. Finally, based on the research highlighted above, the way forward of C-RAN evolution is discussed.
