Large-scale Computer Systems Research Articles

Parallel clustering algorithms for structured AMR

We compare several different parallel implementation approaches for the clustering operations performed during adaptive meshing operations in patch-based structured adaptive mesh refinement (SAMR) applications. Specifically, we target the clustering algorithm of Berger and Rigoutsos, which is commonly used in many SAMR applications. The baseline for comparison is a single program, multiple data extension of the original algorithm that works well for up to O ( 10 2 ) processors. Our goal is a clustering algorithm for machines of up to O ( 10 5 ) processors, such as the 64K-processor IBM BlueGene/L (BG/L) system. We first present an algorithm that avoids unneeded communications of the baseline approach, improving the clustering speed by up to an order of magnitude. We then present a new task-parallel implementation to further reduce communication wait time, adding another order of magnitude of improvement. The new algorithms exhibit more favorable scaling behavior for our test problems. Performance is evaluated on a number of large-scale parallel computer systems, including a 16K-processor BG/L system.

Study of OpenMP applications on the InfiniBand-based software distributed shared-memory system

For the past decades computer engineers have focused on building high-performance and large-scale computer systems with low-cost. One of the examples is a distributed-memory computer system like a cluster, where fast processing nodes to use commodity processors are connected through a high speed network. But it is not easy to develop applications on this system, because a programmer must consider all data and control dependences between processes and program them explicitly. For alleviating this problem the distributed virtual shared-memory (DVSM) system has been proposed. It is well known that the performance of the DVSM system highly depends on the network’s performance and programming semantics, and currently its performance is very limited on a conventional network. Recently many advanced hardware-based interconnection technologies have been introduced, and one of them is the InfiniBand Architecture (IBA) which supports shared-memory programming semantics by means of remote direct-memory access (RDMA) and atomic operations. In this paper, we present the implementation of our InfiniBand-based DVSM system and analyze the performance of SPEC OMP benchmarks in detail by comparing with the DVSM based on the traditional network architecture and the hardware shared-memory multiprocessor (SMP) system. As experiment result, we show that our DVSM system to use full features of the IBA can improve the performance significantly over the IPoIB-based traditional system on the IBA, and furthermore the performance of one application on the IBA-based DVSM system is better than on the hardware SMP.

Classroom support and system administration support for large‐scale educational computer system

AbstractUniversity and other educational computer systems may have several thousands to several tens of thousands of users, and several hundreds to several thousands of terminals. When the scale of the system is enlarged, not only is the burden of management increased, but various problems are produced in the system configuration and administration due to the character of the educational system, in which many identical manipulations are performed at the same time. This study was performed in the educational computer system of the Osaka University Information Processing Education Center (now the Cybermedia Center). Through the design and development of a system to support smooth administration of the computer system and classroom activities in the center, the functions needed for support were organized and practical problems were identified, together with methods for their solution. The new system was operated at the Osaka University Information Processing Education Center and was found to contribute greatly to classroom and administration support. © 2003 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 86(9): 42–52, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjb.10180

The Design and Development of a Financial Cybermarket with a Bundle Trading Mechanism

:Fundamental changes are taking place in the world’s financial markets that will challenge stock exchanges. As the twenty-first century approaches, more and more exchanges are adopting automated floorless trading mechanisms. This article outlines an attempt to develop an automated financial cybermarket with a bundle-matching and -execution mechanism. The market matches orders in bundles and provides a price-discovery mechanism. In many real-world situations, value is derived from a combination of assets or resources. The value will be significantly reduced when even one asset or resource is excluded from the bundle. The mechanism can be used to trade financial portfolios as well as a combination of commodities. When used for simple asset trading, the bundle mechanism is, in effect, a generalization of the k-double auction. From a computing point of view, the electronic market is a large-scale distributed computing system. Emerging technologies like Java and distributed objects can be used to create electronic markets that can be accessed universally. The technologies support interactive and dynamic information exchanges between traders and market. Computational issues, such as asynchronous processing, currency control, and scalability, are also addressed.

Large scale computing system EMI test philosophy

Large scale computing system EMI test philosophy

Fault-Tolerant Process Control and Some Future Directions

Fault-tolerant means the ability, when one or more than one key parts of system is faulty, the system could continuously, stably, and reliably operate and autonomously keep the specified function, or lower the function in an acceptable variable ranges, by taking corresponding method (Levis, 1987). The research of fault-tolerant technique, is originally come from the simulating to natural life’s activities. Later, for the special requirement of defense and military industry, this technique was firstly applied in large-scale computer systems (Chen, 1981; Spitzer, 1988; Sadehi, 1995). Recent years, as the industrial developed and the systems’ scale became larger and larger, manufacture (or production) safety and operation economics have been paid more and more attentions (Levis, 1987; Makansi, 1995). Thus lead to the higher requirement of reliability and effectiveness of process control systems, by which the safety and economics of manufacturing is ensured. This results in the research and development of fault-tolerant process control.In this paper, we try to give some view and comments to the present state of fault-tolerant techniques used in process control, and propose some future directions for this research.

The impact of I/O on program behavior and parallel scheduling

In this paper we systematically examine various performance issues involved in the coordinated allocation of processor and disk resources in large-scale parallel computer systems. Models are formulated to investigate the I/O and computation behavior of parallel programs and workloads, and to analyze parallel scheduling policies under such workloads. These models are parameterized by measurements of parallel programs, and they are solved via analytic methods and simulation. Our results provide important insights into the performance of parallel applications and resource management strategies when I/O demands are not negligible.

A family history of Honeywell's large-scale computer systems

Honeywell Information Systems marketed a broad line of large computer hardware and software, whose heritage goes back to the early 1960s. The authors of this article played key roles in the evolution of today's software (King) and hardware (Shelly).

A massively parallel fault-tolerant architecture for time-critical computing

Building large-scale parallel computer systems for time-critical applications is a challenging task since the designers of such systems need to consider a number of related factors such as proper support for fault tolerance, efficient task allocation and reallocation strategies, and scalability. In this paper we propose a massively parallel fault-tolerant architecture using hundreds or thousands of processors for critical applications with timing constraints. The proposed architecture is based on an interconnection network called thebisectional network. A bisectional network is isomorphic to a hypercube in that a binary hypercube network can be easily extended as a bisectional network by adding additional links. These additional links add to the network some rich topological properties such as node symmetry, small diameter, small internode distance, and partitionability. The important property of partitioning is exploited to propose a redundant task allocation and a task redistribution strategy under realtime constraints. The system is partitioned into symmetric regions (spheres) such that each sphere has a central control point. The central points, calledfault control points (FCPs), are distributed throughout the entire system in an optimal fashion and provide two-level task redundancy and efficiently redistribute the loads of failed nodes. FCPs are assigned to the processing nodes such that each node is assigned two types of FCPs for storing two redundant copies of every task present at the node. Similarly, the number of nodes assigned to each FCP is the same. For a failure-repair system environment the performance of the proposed system has been evaluated and compared with a hypercube-based system. Simulation results indicate that the proposed system can yield improved performance in the presence of a high number of node failures.

Simulation with active objects: an approach to combined modelling

This paper concerns the combined modelling approach developed for evaluation of performability of large scale computer systems. It presents how to combine queuing networks and Generalised Stochastic Petri Nets in a single model, taking advantages of common abstract features. The main objective of this paper is to present a solution concept involving simulation with active objects that clearly reflects the logic of the simulated objects. Therefore an initial outline of the active objects and their properties, services and structures, which together serve to provide a combined modelling, is given.

The use of a fuzzy decision-making method in a large-scale computer system choice problem

We recently proposed a Fuzzy Decision-Making Method for multiple objectives problems in an uncertain and ill-defined environment. This paper presents a real and large-scale application of the method. The problem is to decide what sort of computer system for research and education should be renewed in the department of computer engineering. One member of a renewal committee appears as a decision-maker. After describing the outline of the Fuzzy Decision-Making Method, the decision analysis process, objective hierarchy, identification of decision-maker's preference structure, rating and ranking of alternatives, and sensitivity analysis is presented.

Preface

Recent advances in semiconductor and interconnection technologies and microprocessor design have resulted in a choice of computing options that has redefined the role of traditional large-scale computer systems. These systems are increasingly used as servers for databases, files, communications, and centralized operations. In addition to their ability to handle large volumes of data, transactions, and computations at high speeds, the systems are distinguished by high levels of data integrity and availability, which are required to reliably process critical applications.

THE COMPUTERIZATION OF SOCIAL SECURITY: THE WAY FORWARD OR A STEP BACKWARDS?

The Department of Social Security is one of the largest government departments, consuming 30 percent of public spending and employing around 10 per cent of central government StJ. Before 1980 there were no computers in its local administrative centres. The Operational Strategy was a plan to change this, involving the construction of largescale computer systems and the installation of 40,000 terminals in 1,000 local offices and unemployment offices. It is the largest programme of computerization ever undertaken in Europe. Full national coverage is expected in 1991. But the project has been plagued with problems: costs have soared from £700m. to £2, CCXh‐n. and rising, planned savings have fallen, expected benefits seem unlikely to be achieved and implementation errors have caught the public eye. What caused these problems and are they as serious as reports in the press might lead the public to believe? Computerization has the potential to make possible many far reaching changes to administrative functions and working practices. It is important to study the decisional implications of these changes before the computers themselves become embedded in the structure of the organization.

Large-scale computer simulation and systems design of computer-integrated manufacturing

It is a well known fact that the design of computer-integrated manufacturing systems is an extremely comprehensive task due to the need to consider the complex as a whole. Based on the results of a case study, it is suggested that the system requirements of CIM design creates a need for large-scale simulation, as this kind of simulation is well suited to reflect the complex interactions which are inherent in a CIM environment. The system's definition and the construction of a large-scale simulation model for a particular type of manufacturing operation are discussed. Some experimental results from the model, which reveal certain overall systems behaviours of the manufacturing operation under investigation, are also presented.

Time-dependent rock—Shotcrete interaction a numerical shortcut

Shotcrete is an essential support element for tunnels driven according to the principles of the New Austrian Tunnelling Method (NATM). Therefore the state of stresses and strains prevailing in the shotcrete lining is a major factor for assessing the stability of the opening. The rock-shotcrete interaction near the tunnel face is governed by the non-linear time-dependent material behaviour of shotcrete and the time-dependent excavation process. For a realistic assessment of this problem, large-scale calculations based on the Finite Element Method (FEM) are carried out. The calculation models are 3-dimensional and the time dependence is taken account of by a time stepping procedure. This approach implies extensive CPU time even on large-scale computer systems. The present paper is an attempt to analyze the influence of the essential parameters on the rock-shotcrete interaction at an early age of the shotcrete and to derive a hypothetical modulus of elasticity of the shotcrete lining, which allows an approximative but realistic assessment of the real 3-dimensional rock-shotcrete interaction by means of a 2-dimensional FE-Model without an extensive time-stepping procedure.

Software technology in Taiwan

The steps being taken by Taiwan to shift its focus from manufacturing to information industries are described. To develop manpower, a five-year computer training program called the Software Engineering Institute, to cultivate local software personnel, and the Marketing Information Center have been set up. The government has sponsored four national programs, including the Financial Information System, the Taiwan Meteorological System, the Residential Registration System, and the Patient Care System, that reflect a commitment to the effective use of large-scale computer systems. A number of products have been developed, and a four-year project is underway to develop more. International initiatives and a project to provide local industry a distributed software-development environment are also discussed. Other research is summarized. >

Adaptive storage control for page frame supply in large scale computer systems

A real storage management algorithm called Adaptive Control of Page-frame Supply (ACPS) is described. ACPS employees three strategies: prediction of the demand for real page frames, page replacement based on the prediction, and working set control. Together, these strategies constitute the real page frame allocation method, and contribute to short and stable response times in conversational processing environments. ACPS is experimentally applied to the VOS3 operating system. Evaluation of ACPS on a real machine shows that TSS response times are not affected too strongly by king-size jobs and ACPS is successful in avoiding paging delay and thrashing. ACPS prevents extreme shortages of real storage in almost all cases.

Low dielectric constant multilayer glass-ceramic substrate with Ag-Pd wiring for VLSI package

Glass-ceramic materials with low dielectric constant, which can be sintered at about 900 degrees in air, have been developed. Conductors with extremely low electrical resistivity due to silver and palladium particle shape control have been obtained. The electrical design must be considered for the packaging substrate. Pulse transmission properties are influenced greatly by the kinds of ground planes and wiring planes used. Therefore, basic pulse-transmission properties, such as propagation delay, characteristic impedance, and crosstalk coupling noise, were measured for various multilayer structures. Technologies that allow precise control of substrate properties have been established. Characteristics for the multilayer glass-ceramic (MGC) substrate are summarized. It is shown that the low-dielectric-constant MGC substrate, with silver-palladium conductors, can be applied to VLSI multichip packaging substrate for use in large-scale computer systems. >

Turbo Rand: Monte Carlo sampling and simulations

The popularity of statistical simulation for a wide variety of quantitative investigations has created considerable interest in the generation of pseudo random variables. While the most statistical software packages contain procedures for generating random variates, even those for large scale computer systems are commonly limited to the production of a very few types. The Turbo Rand software provides a welcome addition for at least one computer usage group.

Planning for the Solid-State Millennium

The most common performance bottleneck in today's large-scale commercial computer systems is disk I/O. And the problem is worsening with each new generation of magnetic disk. This article identifies the trends that made this situation inevitable and explains why the next performance revolution will not be the introduction of faster CPUs or exotic new parallel architectures but the widespread use of technology that is already available – semiconductor-based mass storage.