Cluster Of Processors Research Articles

Fast and accurate method for flexible ligand superposition and shapebased screening

The core algorithm is based on the alignment of a set of optimal atom triads followed by volume overlap scoring. The method can process approximately 1000 conformations per second on a modern computer. For flexible superposition we demonstrate on a large data set of crystal structures how accurate alignments can be obtained rapidly (less than one second per molecule) in an automated fashion. Next, we apply the method to virtual screening and show high enrichment rates across a broad range of targets and ligands. A one million compound database (100 conformations per compound) can be processed in approximately 15 minutes on a 100 processor cluster, making this method attractive for pre-screening large databases before downstream pharmacophore-based or docking screens. from 4th German Conference on Chemoinformatics Goslar, Germany. 9–11 November 2008

JChainsAnalyser: an ImageJ-based stand-alone application for the study of magneto-rheological fluids

JChainsAnalyser is a Java-based program for the analysis of two-dimensional images of magneto-rheological fluids (MRF) at low concentration of particles obtained using the video-microscopy technique. MRF are colloidal dispersions of micron-sized polarizable particles in a carrier fluid with medium to low viscosity. When a magnetic field is applied to the suspension, the particles aggregate forming chains or clusters. Aggregation dynamics [P. Domínguez-García, S. Melle, J.M. Pastor, M.A. Rubio, Phys. Rev. E 76 (2007) 051403] and morphology of the aggregates [P. Domínguez-García, S. Melle, M.A. Rubio, J. Colloid Interface Sci. 333 (2009) 221–229] have been studied capturing images of the fluid and analyzing them by using this software. The program allows to analyze automatically the MRF images by means of an adequate combination of different imaging methods, while magnitudes and statistics are calculated and saved in data files. It is possible to run the program on a desktop computer, using the GUI (graphical user interface), or in a cluster of processors or remote computer by means of command-line instructions. Program summary Program title: JChainsAnalyser Catalogue identifier: AEDT_v1_0 Program summary URL: http://cpc.cs.qub.ac.uk/summaries/AEDT_v1_0.html Program obtainable from: CPC Program Library, Queen's University, Belfast, N. Ireland Licensing provisions: Standard CPC license, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines in distributed program, including test data, etc.: 79 071 No. of bytes in distributed program, including test data, etc.: 4 367 909 Distribution format: tar.gz Programming language: Java 2 Computer: Any computer with Java Runtime Environment (JRE) installed Operating system: Any OS with Java Runtime Environment (JRE) installed RAM: Typically, 3.3 MB Classification: 23 External routines: ImageJ, ij-imageIO, jdom, L2FProd Nature of problem: The video-microscopy technique usually produces quite a big quantity of images to analyze. Although ImageJ gives the required filters and methods for image analysis, it fails when a large number of images is used. Moreover, an adequate combination of filters is needed for the segmentation and binarization of this kind of images. Solution method: JChainsAnalyser filters and analyses any quantity of MRF images automatically, so the application can be run on a desktop computer or using a cluster of processors. It can be run in a desktop computer using the GUI (graphical user interface) or by a command-line interface. JChainsAnalyser uses XML files to define input/output data and Java to ensure portability between operating systems. It also utilizes an image algorithm based on the application of different and adaptative ImageJ's filters. Running time: The test run provided takes only a few seconds.

THE LOOMING PETASCALE

“IN THE LATE 1990s, all supercomputing machines looked alike,” according to University of Utah chemistry professor Thomas E. Cheatham III. In general, they had similar structures—clusters of processors that communicate with fast networks. Now, that’s all about to change as a new generation of ultrafast supercomputers looms on the horizon. As a result, computational chemistry is facing the need to radically retool its methods. The new petascale computers will be 1,000 times faster than the terascale supercomputers of today, performing more than 1,000 trillion operations per second. And instead of machines with thousands of processors, petascale machines will have many hundreds of thousands that simultaneously process streams of information. Not everyone is building these computers in the same way. Some make use of the massively parallel graphical processing units found in high-end video game consoles, some are packing more computing core elements into single chips than are found in conventional chips. “The...

Safe and Distributed Kinodynamic Replanning for Vehicular Networks

This work deals with the problem of planning collision-free motions for multiple communicating vehicles that operate in the same, partially-observable environment in real-time. A challenging aspect of this problem is how to utilize communication so that vehicles do not reach states from which collisions cannot be avoided due to second-order motion constraints. This paper initially shows how it is possible to provide theoretical safety guarantees with a priority-based coordination scheme. Safety means avoiding collisions with obstacles and between vehicles. This notion is also extended to include the retainment of a communication network when the vehicles operate as a networked team. The paper then progresses to extend this safety framework into a fully distributed communication protocol for real-time planning. The proposed algorithm integrates sampling-based motion planners with message-passing protocols for distributed constraint optimization. Each vehicle uses the motion planner to generate candidate feasible trajectories and the message-passing protocol for selecting a safe and compatible trajectory. The existence of such trajectories is guaranteed by the overall approach. The theoretical results have also been experimentally confirmed with a distributed simulator built on a cluster of processors and using applications such as coordinated exploration. Furthermore, experiments show that the distributed protocol has better scalability properties when compared against the priority-based scheme.

EXPERIENCES WITH MESH-LIKE COMPUTATIONS USING PREDICTION BINARY TREES ∗

In this paper we aim at exploiting the temporal coherence among successive phases of a computation, in order to implement a load-balancing technique in mesh-like computations to be mapped on a cluster of processors. A key concept, on which the load balancing schema is built on, is the use of a Predictor component that is in charge of providing an estimation of the unbalancing between successive phases. By using this information, our method partitions the computation in balanced tasks through the Prediction Binary Tree (PBT). At each new phase, current PBT is updated by using previous phase computing time for each task as next-phase's cost estimate. The PBT is designed so that it balances the load across the tasks as well as reduces dependency among processors for higher performances. Reducing dependency is obtained by using rectangular tiles of the mesh, of almost-square shape (i. e. one dimension is at most twice the other). By reducing dependency, one can reduce inter-processors communication or exploit local dependencies among tasks (such as data locality). Furthermore, we also provide two heuristics which take advantage of data-locality. Our strategy has been assessed on a significant problem, Parallel Ray Tracing. Our implementation shows a good scalability, and improves performance in both cheaper commodity cluster and high performance clusters with low latency networks. We report different measurements showing that tasks granularity is a key point for the performances of our decomposition/mapping strategy.

Parallel and Distributed Computing Techniques, Selection of papers from ISPDC 2008

Parallel and Distributed Computing Techniques, Selection of papers from ISPDC 2008

Parallel solution of large-scale algebraic Bernoulli equations with the matrix sign function method

We investigate the numerical solution of algebraic Bernoulli equations (ABE) via the Newton iteration for the matrix sign function. Bernoulli equations are nonlinear matrix equations arising in control and systems theory in the context of stabilisation of linear systems, coprime factorisation of rational matrix-valued functions, as well as model reduction. The algorithm proposed here is easily parallelisable and thus provides an efficient tool to solve large-scale problems. We report the parallel performance and scalability of our parallel implementations on a cluster of Intel Xeon processors.

Performance Analysis for Large-Scale Parallel Microscopic Traffic Simulation System

PMTS (Parallel Microscopic Traffic Simulation) is a large-scale microscopic traffic network simulation system using a real traffic network of Shanghai, China. It describes traffic events in object oriented mode, uses MPI (Message Passing Interface) as communicate middleware, partitions the simulation traffic network into subnetworks, and runs simulations of subnetworks concurrently on a cluster of processors connected by high-speed Ethernet. To optimize the performance of PMTS, a number of challenges need to be solved, including traffic network partition, subnetworks communication and synchronization, and workload balance. The parallelization of PMTS is domain decomposition, which means that the geographical region for simulation is decomposed into several domains of similar size and each processor of the PC cluster is responsible for a different geographical area. The partition method needs to fulfill two conditions to make it efficient: minimize the communication overhead and partition the subnetworks of the equal computation load. The goals of maximizing the parallelization and minimizing the communication overhead is conflict, and the tradeoff should be made. PVTM simulates the vehicles activities on the traffic network, and the computation overhead for one vehicle is certain. The time overhead on one processor depends on the number of vehicles for simulation. The individual simulation of vehicle activity includes: vehicle generation, vehicle running, and the vehicle going through the boundary zone. The derivation of the predictive performance is demonstrated and the calculation of the time cost for one processor with a certain number of vehicles is provided. In order to be efficient, the load on different processors should be as similar as possible. The load depends on the actual vehicle number on the respective processor that also varies as the simulation run. In each time step, the workload is collected for each processor. For the next iteration, the information will be fed back to the load-balancer, which will move the boundary of the grid to balance the vehicle numbers among the domains. The performance metrics of the simulation system mainly depend on the load balance, scalability, speedup, and efficiency. The performance optimization of PMTS has been proved to be effective when it is put into use and the experiment results are also provided which match the analysis.

Solving linear-quadratic optimal control problems on parallel computers

We discuss a parallel library of efficient algorithms for the solution of linear-quadratic optimal control problems involving large-scale systems with state-space dimension up to 𝒪(104). We survey the numerical algorithms underlying the implementation of the chosen optimal control methods. The approaches considered here are based on invariant and deflating subspace techniques, and avoid the explicit solution of the associated algebraic Riccati equations in case of possible ill-conditioning. Still, our algorithms can also optionally compute the Riccati solution. The major computational task of finding spectral projectors onto the required invariant or deflating subspaces is implemented using iterative schemes for the sign and disk functions. Experimental results report the numerical accuracy and the parallel performance of our approach on a cluster of Intel Itanium-2 processors.

Cooperation in multi‐organization scheduling

AbstractThe distributed nature of the grid results in the problem of scheduling parallel jobs produced by several independent organizations that have partial control over the system. We consider systems in which each organization owns a cluster of processors. Each organization wants its tasks to be completed as soon as possible. In this paper, we model an off‐line system consisting of N identical clusters of m processors. We show that it is always possible to produce a collaborative solution that respects participants' selfish goals, at the same time improving the global performance of the system. We propose an algorithm (called MOLBA) with a guaranteed worst‐case performance ratio on the global makespan, equal to 4. Next, we show that a better bound (equal to 3) can be obtained in a specific case when the last completed job requires at most m/2 processors. Then, we derive another algorithm (called ILBA) that in practice improves the proposed, guaranteed solution by further balancing the schedules. Finally, by an extensive evaluation by simulation, we show that the algorithms are efficient on typical instances. Copyright © 2008 John Wiley & Sons, Ltd.

Performance Study of GasTurbnLab, an Agent-Based Multi-Physics Problem Solving Environment for the Gas Turbine Engine Simulation

Multiphysics applications are real world problems with a large number of different shape components that obey different physical laws and manufacturing constraints and interact with each other through geometric and physical interfaces. They demand accurate and efficient solutions and a modern type of computational modeling, which designs the whole physical system with as much detail as possible. The simulation of gas turbine engine is such a multiphysics application and is realized with GasTurbnLab, an agent-based Multiphysics Problem Solving Environment (MPSE). Its performance and evaluation study is presented in this paper. For this, a short description of the software components and hardware infrastructure is given. The performance and the scalability of the parallelism are depicted, and the communication overhead between agents is studied with respect to the number of agents and their location in the “computational grid.” The execution time is recorded, and its analysis verifies the complexity of the solvers in use and the performance of the available hardware. Three different clusters of INTEL Pentium processors were used for experimentation to study how the communication time was affected by processor’s homogeneity/heterogeneity and the different connections between the processors. The study of the numerical experiments shows that the domain decomposition and interface relaxation methodology, along with the usage of agent platforms, does not increase the complexity of the simulation problem, and the communication cost is too low, compared with the computations, to reflect on the total simulation time. Therefore, GasTurbnLab is an efficient example of a complex physical phenomena simulation.

Implementation and evaluation of a microthread architecture

Future many-core processor systems require scalable solutions that conventional architectures currently do not provide. This paper presents a novel architecture that demonstrates the required scalability. It is based on a model of computation developed in the AETHER project to provide a safe and composable approach to concurrent programming. The model supports a dynamic approach to concurrency that enables self-adaptivity in any environment so the model is quite general. It is implemented here in the instruction set of a dynamically scheduled RISC processor and many such processors form a microgrid. Binary compatibility over arbitrary clusters of such processors and an inherent scalability in both area and performance with concurrency exploited make this a very promising development for the era of many-core chips. This paper introduces the model, the processor and chip architecture and its emulation on a range of computational kernels. It also estimates the area of the structures required to support this model in silicon.

An architecture for attacking the communication bottleneck in P systems

The distributed implementation of P systems on a cluster of processors has met with a bottleneck communications problem. When the number of membranes grows in the system, the network gets congested and the time taken to execute an evolution step degrades. In this article, we suggest a software architecture denominated “partially parallel evolution with partially parallel communication”, where some membranes are located in each processor, proxies are used to communicate with membranes located in different processors, and a policy of access control to the communications network is mandatory. With all this, we get a certain parallelism in the system and an acceptable functioning in communications. In addition to this, it establishes a series of equations that allows us to determine in the architecture the optimum number of processors needed, the time required to execute an evolution step, the number of membranes to be located in each processor, and the conditions to determine when it is best to use the distributed solution or the sequential one.

Coarse grained parallel algorithms for graph matching

Parallel graph algorithm design is a very well studied topic. Many results have been presented for the PRAM model. However, these algorithms are inherently fine grained and experiments show that PRAM algorithms do often not achieve the expected speedup on real machines because of large message overheads. In this paper, we present coarse grained parallel graph algorithms with small message overheads that solve the following standard graph problems related to graph matching: finding maximum matchings in convex bipartite graphs, and finding maximum weight matchings in trees. To our knowledge, these are the first efficient parallel algorithms for these problems that are designed for standard commercial parallel machines such as off-the-shelf processor clusters.

Phospholemman Transmembrane Structure Reveals Potential Interactions with Na+/K+-ATPase

Phospholemman (PLM) is a 72-residue bitopic cardiac transmembrane protein, which acts as a modulator of the Na(+)/K(+)-ATPase and the Na(+)/Ca(2+) exchanger and possibly forms taurine channels in nonheart tissue. This work presents a high resolution structural model obtained from a combination of site-specific infrared spectroscopy and experimentally constrained high throughput molecular dynamics (MD) simulations. Altogether, 37 experimental constraints, including nine long range orientational constraints, have been used during MD simulations in an explicit lipid bilayer/water system. The resulting tetrameric alpha-helical bundle has an average helix tilt of 7.3 degrees and a crossing angle close to 0 degrees . It does not reveal a hydrophilic pore, but instead strong interactions between various residues occlude any pore. The helix-helix packing is unusual, with Gly(19) and Gly(20) pointing to the outside of the helical bundle, facilitating potential interaction with other transmembrane proteins, thus providing a structural basis for the modulatory effect of PLM on the Na(+)/K(+)-ATPase. A two-stage model of interaction between PLM and the Na(+)/K(+)-ATPase is discussed involving PLM-ATPase interaction and subsequent formation of an unstable PLM trimer, which readily interacts with surrounding ATPase molecules. Further unconstrained MD simulations identified other packing models of PLM, one of which could potentially undergo a conformational transition to an open pore.

Design and implementation of a massively parallel version of DIRECT

This paper describes several massively parallel implementations for a global search algorithm DIRECT. Two parallel schemes take different approaches to address DIRECT’s design challenges imposed by memory requirements and data dependency. Three design aspects in topology, data structures, and task allocation are compared in detail. The goal is to analytically investigate the strengths and weaknesses of these parallel schemes, identify several key sources of inefficiency, and experimentally evaluate a number of improvements in the latest parallel DIRECT implementation. The performance studies demonstrate improved data structure efficiency and load balancing on a 2200 processor cluster.

A Fully Implicit, Compositional, Parallel Simulator for IOR Processes in Fractured Reservoirs

Summary Naturally fractured reservoirs contain a significant amount of world oil reserves. A number of these reservoirs contain several billion barrels of oil. Accurate and efficient reservoir simulation of naturally fractured reservoirs is one of the most important, challenging, and computationally intensive problems in reservoir engineering. Parallel reservoir simulators developed for naturally fractured reservoirs can effectively address the computational problem. A new accurate parallel simulator for large-scale naturally fractured reservoirs, capable of modeling fluid flow in both rock matrix and fractures, has been developed. The simulator is a parallel, 3D, fully implicit, equation-of-state compositional model that solves very large, sparse linear systems arising from discretization of the governing partial differential equations. A generalized dual-porosity model, the multiple-interacting-continua (MINC), has been implemented in this simulator. The matrix blocks are dis-cretized into subgrids in both horizontal and vertical directions to offer a more accurate transient flow description in matrix blocks. We believe this implementation has led to a unique and powerful reservoir simulator that can be used by small and large oil producers to help them in the design and prediction of complex gas and waterflooding processes on their desktops or a cluster of computers. Some features of this simulator, such as modeling both gas and water processes and the ability of 2D matrix subgridding are not available in any commercial simulator to the best of our knowledge. The code was developed on a cluster of processors, which has proven to be a very efficient and convenient resource for developing parallel programs. The results were successfully verified against analytical solutions and commercial simulators (ECLIPSE and GEM). Excellent results were achieved for a variety of reservoir case studies. Applications of this model for several IOR processes (including gas and water injection) are demonstrated. Results from using the simulator on a cluster of processors are also presented. Excellent speedup ratios were obtained.

Speech recognition systems on the Cell Broadband Engine processor

In this paper we describe our design, implementation, and initial results of a prototype connected-phoneme-based speech recognition system on the Cell Broadband Engine™ (Cell/B.E.) processor. Automated speech recognition decodes speech samples into plaintext (other representations are possible) and must process samples at real-time rates. Fortunately, the computational tasks involved in this pipeline are highly data parallel and can receive significant hardware acceleration from vector-streaming architectures such as the Cell/B.E. Architecture. Identifying and exploiting these parallelism opportunities is challenging and critical to improving system performance. From our initial performance timings, we observed that a single Cell/B.E. processor can recognize speech from thousands of simultaneous voice channels in real time-a channel density that is orders of magnitude greater than the capacity of existing software speech recognizers based on CPUs (central processing units). This result emphasizes the potential for Cell/B.E. processor-based speech recognition and will likely lead to the development of production speech systems using Cell/B.E. processor clusters.

Collective communication: theory, practice, and experience

AbstractWe discuss the design and high‐performance implementation of collective communications operations on distributed‐memory computer architectures. Using a combination of known techniques (many of which were first proposed in the 1980s and early 1990s) along with careful exploitation of communication modes supported by MPI, we have developed implementations that have improved performance in most situations compared to those currently supported by public domain implementations of MPI such as MPICH. Performance results from a large Intel Xeon/Pentium 4 (R) processor cluster are included. Copyright © 2007 John Wiley & Sons, Ltd.

Scheduling Policies for Processor Coallocation in Multicluster Systems

Building multicluster systems out of multiple, geographically distributed clusters interconnected by high-speed wide-area networks can provide access to a larger computational power and to a wider range of resources. Jobs running on multiclusters and, more generally, in grids, may require (processor) coallocation, i.e., the simultaneous allocation of resources (processors) in different clusters or subsystems of a grid. In this paper, we propose four scheduling policies for processor coallocation in multiclusters, and we assess with simulations their performance under a wide variety of parameter settings. In particular, in our simulations we use synthetic workloads and workloads derived from the logs of actual systems and from runtime measurements. We conclude that although coallocation makes scheduling more difficult and the wide-area communication critically impacts the performance, there is a wide range of realistic applications that may benefit from coallocation. However, unrestricted coallocation is not recommended: Limiting the total job size or the number or the sizes of their components improves performance.