Performance evaluation of the SX‐6 vector architecture for scientific computations

The growing gap between sustained and peak performance for scientific applications is a well‐known problem in high‐performance computing. The recent development of parallel vector systems offers the potential to reduce this gap for many computational science codes and deliver a substantial increase in computing capabilities. This paper examines the intranode performance of the NEC SX‐6 vector processor, and compares it against the cache‐based IBM Power3 and Power4 superscalar architectures, across a number of key scientific computing areas. First, we present the performance of a microbenchmark suite that examines many low‐level machine characteristics. Next, we study the behavior of the NAS Parallel Benchmarks. Finally, we evaluate the performance of several scientific computing codes. Overall results demonstrate that the SX‐6 achieves high performance on a large fraction of our application suite and often significantly outperforms the cache‐based architectures. However, certain classes of applications are not easily amenable to vectorization and would require extensive algorithm and implementation reengineering to utilize the SX‐6 effectively. Copyright © 2005 John Wiley & Sons, Ltd.     

Reversing course: Issues in modeling legacy systems

The Enhanced Traffic Management System (ETMS) is a legacy system used by the Federal Aviation Administration (FAA) to support air traffic flow management. Air traffic flow management is the strategic control of air traffic to minimize delays and congestion and maximize the throughput of aircraft throughout the National Airspace System (NAS). This paper discusses the reasons for modeling a legacy system, problems and advantages encountered in modeling an operational system, and describes the construction of a simulation model of ETMS. Originally written in Pascal to run on Apollo workstations under the Aegis Domain operating system, ETMS has been converted to C/C++ and ported to HP servers and workstations running HP-UX, a POSIX-compliant version of UNIX. The objectives of the modeling task were to assess performance of the ported system and to provide a basis for evaluating a possible redesign/re-architecture of the system. The initial plan was to develop one or two models aimed at the network aspects (both LAN and WAN) of ETMS at a relatively high level, and then to develop a more detailed model to look at specific workstation/server issues.As is shown in this paper, issues of existing system design and documentation and the availability (or the lack) of data continually arose. Nevertheless, a reasonable set of working assumptions were derived which allowed modeling and evaluation to proceed. Thus, the quantitative and qualitative results obtained provided information and lessons learned that can be built upon. Moreover, the second of the stated goals (to provide a basis for possible redesign) was also achieved because there is now a baseline for future design/architecture studies. The focus of this paper is to provide insights into the issues involved in modeling an existing system rather than the results of the model itself.     

DAFS实现方法及其在NAS集群中的应用

DAFS是目前新出现的一种文件访问和管理协议。该文介绍了DAFS协议的基本特点,分析了DAFS的实现过程及其实现方法,最后讨论了DAFS在NAS集群技术中的应用。     

一种基于磁盘调度的在线数据重构算法

作为一个完善的卷管理器的设计,在线数据重构机制是必不可少的。数据重构算法既可以以带区为导向,也可以以磁盘为导向。该文在分析了基于带区调度的重构算法后,给出了一种基于磁盘调度的重构算法。实验结果表明,基于磁盘调度的重构算法能更有效地利用系统资源。     

基于自适应学习率优化的AdaNet改进

人工神经网络的自适应结构学习（AdaNet）是基于Boosting集成学习的神经结构搜索框架,可通过集成子网创建高质量的模型。现有的AdaNet所产生的子网之间的差异性不显著,因而限制了集成学习中泛化误差的降低。在AdaNet设置子网网络权重和集成子网的两个步骤中,使用Adagrad、RMSProp、Adam、RAdam等自适应学习率方法来改进现有AdaNet中的优化算法。改进后的优化算法能够为不同维度参数提供不同程度的学习率缩放,得到更分散的权重分布,以增加AdaNet产生子网的多样性,从而降低集成学习的泛化误差。实验结果表明,在MNIST（Mixed National Institute of Standards and Technology database）、Fashion-MNIST、带高斯噪声的Fashion-MNIST这三个数据集上,改进后的优化算法能提升AdaNet的搜索速度,而且该方法产生的更加多样性的子网能提升集成模型的性能。在F1值这一评估模型性能的指标上,改进后的方法相较于原方法,在三种数据集上的最大提升幅度分别为0.28%、1.05%和1.10%。     

唐钢薄板坯连铸和超薄带钢液压系统的清洗

以唐钢薄板坯连铸和超薄带钢工程液压系统的化学清洗、油冲洗为例，说明液压系统的工作原理、在油冲洗前进行化学清洗的必要性及化学清洗、油冲洗施工方案的合理性。     

Multi-core optimization for conjugate gradient benchmark on heterogeneous processors

Developing parallel applications on heterogeneous processors is facing the challenges of ‘memory wall’, due to limited capacity of local storage, limited bandwidth and long latency for memory access. Aiming at this problem, a parallelization approach was proposed with six memory optimization schemes for CG, four schemes of them aiming at all kinds of sparse matrix-vector multiplication (SPMV) operation. Conducted on IBM QS20, the parallelization approach can reach up to 21 and 133 times speedups with size A and B, respectively, compared with single power processor element. Finally, the conclusion is drawn that the peak bandwidth of memory access on Cell BE can be obtained in SPMV, simple computation is more efficient on heterogeneous processors and loop-unrolling can hide local storage access latency while executing scalar operation on SIMD cores.     

电视台制播系统中存储方案的设计

磁盘存储技术已在电视台制播系统中获得大量应用.本文简要介绍并比较了各种存储方式:DAS、NAS、FC-SAN、ISCSI-SAN等.     

基于大容量镜像NAS存储结构的设计

针对IP SAN网络存储结构的部署与维护,在综合考虑柳州广播电视台业务发展和实际需求的基础上,设计了基于大容量镜像的NAS存储结构,结果证明,新的NAS存储结构简单,成本低,易于维护,故障排查和处理更迅速,扩展性较高,并且实现了实时同步与快速切换的主备高可用存储,确保了节目流程中数据的安全性。     

428XL仪器的NAS盘的应用与维修

NAS盘的使用提高了野外地震勘探的生产效率。文章介绍了NAS的构成以及安装和使用方法,并讨论了在使用过程中常见问题的解决方法。