一种基于实测的高维动态负载平衡方法

针对大规模科学计算中的强非规则结构负载问题,作者开发出一种基于实测的动态负载平衡方法．首先,将由规则结构化网格组成的模拟区域剖分成多块;其次,把块的高维坐标转换成一维Hilbert空间填充曲线（HSFC）索引;然后,基于实测信息采用多层均权法剖分按一维HSFC索引排列的块;最后根据剖分信息重分配块以平衡负载．它把仅适用于一维的多层均权法扩展到二维和三维,并引入更多的实测信息和块数据结构．与ISP方法相比,该方法在64个CPU上提高负载平衡效率10％,在某MPP的500个CPU上模拟强非规则结构负载问题时,获得了88％的负载平衡效率和84％的并行效率．     

Characterization of database access pattern for analytic prediction of buffer hit probability

The analytic prediction of buffer hit probability, based on the characterization of database accesses from real reference traces, is extremely useful for workload management and system capacity planning. The knowledge can be helpful for proper allocation of buffer space to various database relations, as well as for the management of buffer space for a mixed transaction and query environment. Access characterization can also be used to predict the buffer invalidation effect in a multi-node environment which, in turn, can influence transaction routing strategies. However, it is a challenge to characterize the database access pattern of a real workload reference trace in a simple manner that can easily be used to compute buffer hit probability. In this article, we use a characterization method that distinguishes three types of access patterns from a trace: (1) locality within a transaction, (2) random accesses by transactions, and (3) sequential accesses by long queries. We then propose a concise way to characterize the access skew across randomly accessed pages by logically grouping the large number of data pages into a small number of partitions such that the frequency of accessing each page within a partition can be treated as equal. Based on this approach, we present a recursive binary partitioning algorithm that can infer the access skew characterization from the buffer hit probabilities for a subset of the buffer sizes. We validate the buffer hit predictions for single and multiple node systems using production database traces. We further show that the proposed approach can predict the buffer hit probability of a composite workload from those of its component files.     

The available capacity of a privately owned workstation environment

Powerful workstations interconnected by networks have become widely available as sources of computing cycles. Each workstation is typically owned by a single user in order to provide a high quality of service for the owner. In most cases, an owner does not have computing demands as large as the capacity of the workstation. Therefore, most of the workstations are underutilized. Nevertheless, some users have demands that exceed the capacities of their workstations. In order to effectively share the capacity of workstations, there must be algorithms that allocate available capacity and long periods when owners do not use their stations. To understand the profile of station availability, we analyzed the usage patterns of a cluster of workstations. The workstations were available more than 75% of the time observed. Large capacities were steadily available on an hour to hour, day to day, and month to month basis. These capacities were available not only during the evening hours and on weekends, but during the busiest times of normal working hours. A stochastic model was developed which was based on an analysis of the relative frequency distribution and the correlation of available and non-available interval lengths. A 3-stage hyperexponential cumulative distribution has been fitted to the observed cumulative relative frequency of available periods. The fitted distribution closely matches the observed relative frequency distribution. This stochastic model is important as a workload generator for the performance evaluation of capacity sharing strategies of a cluster of workstations. The model assists in the design of resource management algorithms that take advantage of the characteristics of the usage patterns.     

Experiences in benchmarking the three supercomputers CRAY-1M, CRAY-X/MP, FUJITSU VP-200 compared with the CYBER 76

The speed of the three supercomputers CRAY-1M, CRAY-X/MP, FUJITSU VP-200 is measured several times. There exist technical numbers like cycle time, start-up times, etc. and numbers for the speed of basic arithmetic operations depending on the vector length, for kernel programs and for a few special production programs. In this article some numbers and some experiences are given for a broader program set, real production programs from a heterogenous workload, typical for a university computer center environment with technical oriented research problems. The intention was to measure the relative speed to the CONTROL DATA CYBER 76, the main computer which the Regionales Rechenzentrum für Niedersachsen (RRZN) at the University of Hannover has operated for more than 10 years. For replacing this computer some investigations were necessary for benchmarking some new computers and supercomputers. Experiences are given with the migration of real programs to the supercomputers and a used benchmark is described. The measured speed factors are given for the three supercomputers compared with the CYBER 76. Very remarkable is the big range of the different speed factors. Some global thoughts about benchmarking, the interpretation of the results for the used benchmark and some special programs with their effects are discussed.     

高速网络环境中的科学计算可视化

可视化技术已经极大地提高了研究人员分析、理解数据的能力，科学计算可视化和高速网络的紧密结合为超级计算提供了一种有效环境。本文主要介绍科学计算可视化在网络环境中所面临的一些问题，提出一种解决负载分配问题的有效方法，最后，总结未来超级计算所需可视化环境的研究工作。     

基于利用率和负载均衡的多核实时调度算法研究

针对分区调度算法在实时多处理器系统中处理器利用率不高的现象,提出一种基于利用率和负载均衡的分区调度算法BUWBPA(Based on Utilization and Workload Balance Partition Algorithm)。该算法在满足任务实时性要求的基础上,以寻求高利用率和负载均衡为目标进行任务分配,将任务分配分成两个阶段:第一个阶段以高利用率为原则,选择任务集内利用率最高的任务先分配;第二个阶段以负载均衡为原则,根据处理器数选择利用率总和等于1或接近于1的任务进行分配,并且在此阶段对于未达到充分利用的处理器,选取可能调度的零星任务,对任务进行再次重新分配,以达到负载均衡和系统最大利用率。实验证明,该算法在实现最大利用率的前提下能很好地达到负载均衡。     

基于脑电“微状态”的驾驶员脑负荷研究

驾驶疲劳影响驾驶员的警觉度和注意能力,同时严重影响道路交通安全。该文在实验室驾驶平台中模拟疲劳驾驶的基础上,提出了一种评估驾驶过程中大脑工作负荷的方法。实验结果表明,当大脑工作负荷较小时,在theta频段左前额高电势微状态出现的频率升高;当大脑工作负荷较大时频率降低。以此为指标,可以对驾驶员的放松和警觉状态进行检测。     

Workload characterization using the CRAY hardware performance monitor

This paper describes some preliminary results about a workload characterization study at a national supercomputer center. This study is part of a larger project to develop an analytic methodology for the construction of arepresentative benchmark set. Our study reports on 292,254 user jobs during the period from June 1992 to July 1993 and represents about one-half of the wall clock time available on the four-processor CRAY Y-MP at the National Center for Supercomputing Applications during this time. The performance statistics we gathered in some sensedefine the workload because of the large fraction of the available time that was recorded. These statistics will allow us to reverse engineer a set of benchmarks that analytically represents the workload at this site.     

Optimization‐based workload distribution in geographically distributed data centers: A survey

Energy efficiency is a contemporary and challenging issue in geographically distributed data centers. These data centers consume significantly high energy and cast a negative impact on the energy resources and environment. To minimize the energy cost and the environmental impacts, Internet service providers use different approaches such as geographical load balancing (GLB). GLB refers to the placement of data centers in diverse geolocations to exploit variations in electricity prices with the objective to minimize the total energy cost. GLB helps to minimize the overall energy cost, achieve quality of service, and maximize resource utilization in geo‐distributed data centers by employing optimal workload distribution and resource utilization in the real time. In this paper, we summarize various optimization‐based workload distribution strategies and optimization techniques proposed in recent research works based on commonly used optimization factors such as workload type, load balancer, availability of renewable energy, energy storage, and data center server specification in geographically distributed data centers. The survey presents a systemized and a novel taxonomy of workload distribution in data centers. Moreover, we also debate various challenges and open research issues along with their possible solutions.     

图切割快速生成扇区的蚁群算法

扇区划分是平衡管制员工作负荷、提升空域通行能力的有效技术措施。采用Voronoi图自顶向下切割空域的方法具有自动保证扇区凸性、连通性和压缩性的特性,但计算时间过长。根据航迹状态计算工作负荷,构建了Voronoi图自顶向下切割空域模型,设计了动态步长蚁群搜索算法。测试结果表明,在太原高空划分成4个扇区的情况下,与MC-CLFV（Monte Carlo method by changing location of flexible vertices）算法相比较,采用动态步长蚁群搜索算法求解,扇区的不平衡性显著减少,总工作负荷也减少了11.2%。大规模重复实验表明,随扇区数量增加,解的质量差异减少,但解的效率差异却不断增大,划分8个扇区时,动态步长蚁群搜索算法的计算时间统计中位数值仅为MC-CLFV算法的1/10。这说明该算法在一次划分数量少的时候,质量高而且计算时间短,在采用多层规划扇区的时候,应该采用组合数字和较小的划分方案。研究结果为采用Voronoi图直接切割空域形成管制扇区奠定了基础。