By Topic

Parallel and Distributed Computing, 2005. ISPDC 2005. The 4th International Symposium on

Date 4-6 July 2005

Filter Results

Displaying Results 1 - 25 of 52
  • The 4th International Symposium on Parallel and Distributed Computing - Cover

    Page(s): c1
    Save to Project icon | Request Permissions | PDF file iconPDF (183 KB)  
    Freely Available from IEEE
  • The 4th International Symposium on Parallel and Distributed Computing - Title Page

    Page(s): i - iii
    Save to Project icon | Request Permissions | PDF file iconPDF (89 KB)  
    Freely Available from IEEE
  • The 4th International Symposium on Parallel and Distributed Computing - Copyright

    Page(s): iv
    Save to Project icon | Request Permissions | PDF file iconPDF (39 KB)  
    Freely Available from IEEE
  • The 4th International Symposium on Parallel and Distributed Computing - Table of contents

    Page(s): v - ix
    Save to Project icon | Request Permissions | PDF file iconPDF (93 KB)  
    Freely Available from IEEE
  • General Chairman’s Forward

    Page(s): x
    Save to Project icon | Request Permissions | PDF file iconPDF (20 KB)  
    Freely Available from IEEE
  • Committees

    Page(s): xi
    Save to Project icon | Request Permissions | PDF file iconPDF (31 KB)  
    Freely Available from IEEE
  • A New Era in Computing: Moving Services onto Grid

    Page(s): 3
    Save to Project icon | Request Permissions | PDF file iconPDF (76 KB)  
    Freely Available from IEEE
  • New Challenges in Parallel Optimization

    Page(s): 5
    Save to Project icon | Request Permissions | PDF file iconPDF (75 KB)  
    Freely Available from IEEE
  • From Grid Technologies (in 6th of IST RTD 2002-2006) to knowledge Utility

    Page(s): 8
    Save to Project icon | Request Permissions | PDF file iconPDF (75 KB)  
    Freely Available from IEEE
  • E-NEXT: Network of Excellence - Emerging Network Technologies

    Page(s): 9 - 10
    Save to Project icon | Request Permissions | PDF file iconPDF (93 KB)  
    Freely Available from IEEE
  • Task Scheduling for Dynamic SMP Clusters with Communication on the Fly for Bounded Number of Resources

    Page(s): 13 - 20
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (2912 KB) |  | HTML iconHTML  

    The paper presents an algorithm for scheduling parallel tasks in a parallel architecture based on multiple dynamic SMP clusters. Processors can be switched between shared memory modules in the runtime. Memory modules and processors are organized in computational system on chip modules of a fixed size and are inter-connected by a local communication network implemented in a network-on-chip technology (NoC). Processors located in the same SoC module can share their data and communicate using a technique of data transfers on the fly. A number of such SoC modules can be connected using a global interconnection network to form a larger infrastructure. The presented algorithm schedules initial macro dataflow program graph for such an architecture with a given number of SoC modules, considering a fixed size of a module. First, it distributes program graph nodes among processors, assuming no distribution between SoC modules. Then it transforms and schedules computations and communications to use processor switching and read on the fly facilities. Finally, using genetic algorithm, it divides the whole set of processors into subsets of a given size, which then are mapped to separate SoC modules View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Java Byte Code Scheduling Based on the Most-Often-Used-Paths in Programs with Branches

    Page(s): 21 - 27
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (281 KB) |  | HTML iconHTML  

    The paper presents an introductory optimization algorithm, which can be performed before a Java program is executed in a parallel system. Taking a sequential multithreaded version of a Java program as input information, the aim of the parallel program optimization is to determine an initial distribution of objects on virtual machines so as to decrease direct inter-object communication and balance loads of the virtual machines. The object placement optimization algorithm uses a graphical representation of control dependencies and data dependencies among methods in Java programs. These dependencies are discovered by an analysis of program byte code and stored in the form of relevant macro/dataflow graphs. The placement optimization algorithm tries to optimally assign the macro nodes to processors (JVMs) so as to reduce inter-processor communication overheads. The optimization method first does clustering of macro nodes on unlimited number of processors (logical JVMs) to reduce the execution time of the clustered nodes. Next, merging of the assigned clusters is performed to reduce the number of logical JVMs to the number of real processors. The presented approach is supported by a dynamic, online load balancing mechanism, which applies object migration as proposed in the ADAJ (adaptive distributed applications in Java) project View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Performance Effective Task Scheduling Algorithm for Heterogeneous Computing System

    Page(s): 28 - 38
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (261 KB) |  | HTML iconHTML  

    Finding an optimal solution to the problem of scheduling an application modeled by a directed acyclic graph (DAG) onto a distributed system is known to be NP-complete. The complexity of the problem increases when task scheduling is to be done in a heterogeneous computing system, where the processors in the network may not be identical and take different amounts of time to execute the same task. This paper introduces a performance effective task scheduling (PETS) algorithm for network of heterogeneous system, with complexity O(v+e) (p+log v), which provides optimal results for applications represented by DAGs. The performance of the algorithm is illustrated by comparing the schedule length, speedup, efficiency and the scheduling time with existing algorithms such as, heterogeneous earliest finish time (HEFT) and critical-path on a processor (CPOP) and levelized min time (LMT) reported in this paper. The comparison study based on both randomly generated graphs and graphs of some real applications shows that PETS algorithm substantially outperforms existing algorithms View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Cellular Nonlinear Network Grids - a new Grid Computing Approach

    Page(s): 39 - 42
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (279 KB) |  | HTML iconHTML  

    Throughout this paper a catenation between the universal paradigm of cellular nonlinear networks (CNN) and the innovative approach of grid computing is given. CNN are a massive parallel solution for solving non-linear problems, modelling complex phenomena in medicine, physics and data analysis as well as for powerful image processing and recognition systems. They usually are simulated on local computer systems or built as dedicated VLSI-implementations. However, the research of complex CNN structures and settings require massive computing power and thus can benefit from multi-system open architectures which can be provided by the grid approach. Propositions of two different realizations with grid architecture in mind are given by introducing an algorithm of implementing such methods in a CNN software simulator. First a brief introduction to CNN is given. Afterwards, problems for the current determination of such networks are discussed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Towards Scalable Grid Replica Optimization Framework

    Page(s): 43 - 50
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (219 KB) |  | HTML iconHTML  

    Data grids are becoming a popular infrastructure for distributed, data intensive scientific application. Data access optimization is one of the key features of such systems, as the transfer and access to the huge amounts of data are often the bottleneck of the data intensive tasks. Concept of replication, creation of multiple copies of a data source across multiple grid nodes, was adopted by grid community to increase data availability. In this paper, we present the ongoing work on the grid replica optimization framework that addresses both short-term and long-term replica optimization problems. We describe high level architecture of the optimization framework and discuss the promising results achieved by prototype implementation of several system components View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Comparative Simulation Analysis of P2P System Architectures

    Page(s): 51 - 57
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (362 KB) |  | HTML iconHTML  

    This paper gives a comparative study on multilayer file sharing mechanisms in peer-to-peer (P2P) systems. Based on a well-known P2P system named Gnutella, two system architectures, 1-layer and 2-layer Ngnu architectures are proposed in order to reduce network traffic and to gain better scalability. These architectures can be applied to large-scale systems, such as e-government, e-office and e-library, which need an efficient file-sharing mechanism. To examine the efficiency and effectiveness of the proposed methods, a number of simulations are made with respect to three factors: time-to-lives (TTLs), the number of nodes and the number of queries. To compare these systems, the number of messages and the mean query hit are measured and compared with the original flat Gnutella system View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive window scheduling for a hierarchical agent system

    Page(s): 58 - 65
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (263 KB) |  | HTML iconHTML  

    DIET (distributed interactive engineering toolbox) is a toolbox for the construction of network enabled server (NES) systems. For most NES systems, as for most grid middleware systems, the scheduling system is centralized and can suffer from poor scalability. DIET provides an alternative: low-latency, scalable scheduling services based on a distributed hierarchy of scheduling agents. However, the online scheduling model used currently in DIET can overload interactive servers in high-load conditions and does not allow adaption to task or data dependencies. In this article we consider an alternative model based on active management of the flow of requests throughout the system. We have added support for (1) limiting the number of concurrent requests on interactive servers, (2) server and agent-level queues, and (3) window-based scheduling algorithms whereby the request release rate to servers can be controlled and some re-arrangement of request to host mappings is possible. We present experiments demonstrating that these approaches can improve performance and that the overheads introduced are not significantly different from those of the standard DIET approach View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Enforcing consistency during the adaptation of a parallel component

    Page(s): 66 - 73
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (269 KB) |  | HTML iconHTML  

    As grid architectures provide execution environments that are distributed, parallel and dynamic, applications require to be not only parallel and distributed, but also able to adapt themselves to their execution environment. This article presents a model for designing self-adaptable parallel components that can be assembled to build applications for grid. This model includes the definition of a consistency criterion for the dynamic adaptation of SPMD components. We propose a solution to implement this criterion. It has been evaluated on both synthetic and real codes to exhibit the behavior of the several proposed strategies View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • GMRES Method on Lightweight GRID System

    Page(s): 74 - 82
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (272 KB) |  | HTML iconHTML  

    Grid computing accomplishes high throughput computing by using a very large number of unexploited computing resources. We present a parallel method GMRES to solve large sparse linear systems by the use of a lightweight grid XtremWeb. This global computing platform, just as many popular grid systems, is devoted to multi-parameters generic-applications. We have implemented an important algorithm GMRES which is one of the key methods to resolve large, non-symmetric, linear problems. We discuss the performances of this algorithm deployed on two XtremWeb networks: a local network with 128 non-dedicated PCs in Polytech-Lille of University of Lille I in France, a remote network with 3 clusters of SCGN grid that includes 91 CPUs totally in the High Performance Center for Computer Science of University of Tsukuba in Japan. We compare these performances with those of a MPI implementation of GMRES on the same platform. We present the advantages and shortcomings of our implementation on this lightweight grid XtremWeb View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A Strategyproof Mechanism for Scheduling Divisible Loads in Distributed Systems

    Page(s): 83 - 90
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (318 KB) |  | HTML iconHTML  

    An important scheduling problem is the one in which there are no dependencies between tasks and the tasks can be of arbitrary size. This is known as the divisible load scheduling problem and was studied extensively resulting in a cohesive theory called divisible load theory (DLT). In this paper, we augment the existing divisible load theory with incentives. We develop a strategyproof mechanism for scheduling divisible loads in distributed systems assuming a bus type interconnection and a linear cost model for the processors. The mechanism provides incentives to processors such that it is beneficial for them to report their true processing power and process the assigned load using their full processing capacity. We define the strategyproof mechanism and prove its properties. We simulate and study the implementation of the mechanism on systems characterized by different parameters View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Load Balancing Algorithm in Cluster-based RNA secondary structure Prediction

    Page(s): 91 - 96
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (252 KB) |  | HTML iconHTML  

    RNA secondary structure prediction remains one of the most compelling, yet elusive areas of computational biology. Many computational methods have been proposed in an attempt to predict RNA secondary structures. A popular dynamic programming (DP) algorithm uses a stochastic context-free grammar to model RNA secondary structures, its time complexity is O(N4) and spatial complexity is O(N3 ), where N is the length of sequences. In this paper, a parallel algorithm, which is time-wise and space-wise optimal with respect to the usual sequential DP algorithm, can be implemented using O(N4/P) time and O(N3/P) space in cluster, where P is the number of processors. High efficient utilization of processors and good load balancing are important to the performance of parallel algorithms in cluster systems. Two parallel DP algorithms, which have different mappings of the DP matrix to processors, are evaluated concerning running time. As experiments show, dynamic mapping of DP matrix can achieve better load balancing than the static and improve the efficiency of processors. Thus, the dynamic mapping algorithm is faster and gets better speedups View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Ghost Process: a Sound Basis to Implement Process Duplication, Migration and Checkpoint/Restart in Linux Clusters

    Page(s): 97 - 104
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (507 KB) |  | HTML iconHTML  

    Process management mechanisms (process duplication, migration and checkpoint/restart) are very useful for high performance and high availability in clustering systems. The single system image approach aims at providing a global process management service with mechanisms for process checkpoint, process migration and process duplication. In this context, a common mechanism for process virtualization is highly desirable but traditional operating systems do not provide such a mechanism. This paper presents a kernel service for process virtualization called ghost process, extending the Linux kernel. The ghost process mechanism has been implemented in the Kerrighed single system image based on Linux View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Load Balancing with Migration Based on Synchronizers in PS-GRADE Graphical Tool

    Page(s): 105 - 112
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (478 KB) |  | HTML iconHTML  

    Parallel application control based on application global states is a new concept realized in PS-GRADE. PS-GRADE is a graphical environment for parallel programming, which unifies message passing programming style with control based on global application states. Special processes called synchronizers are responsible for gathering process states, constructing application global states and issuing control signals when necessary to application processes. In this paper, we show how this mechanism can be used as a framework for implementing load balancing with process migration using several methods. With some of these methods, synchronizers collect strongly consistent states of processor loads in the system and workout load balancing decisions including process migration. Asynchronous control signals from synchronizers break current computations and cause a process to migrate onto a less loaded host. Another signal activates a target process and activates restoring of the captured state in it. With other methods we use a special PVM library ynamite PVM instead of a standard version. It extends PVM by checkpointing and process migration with full restoration of the process state View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Measuring and improving quality of parallel application monitoring based on global states

    Page(s): 113 - 120
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (276 KB) |  | HTML iconHTML  

    Effectiveness of parallel/distributed application control based on predicates defined on application global states depends on the performance of an underlying consistent global state monitoring mechanism. Focusing on strongly consistent global states (SCGS) usage, we introduce three measures of SCGS monitoring quality. They reflect delay, state information ageing and state inspection frequency experienced by a monitor. Using the measures, four SCGS monitoring algorithms are compared, including two novel algorithm variants. The comparison is carried out with the use of simulations. The newly introduced algorithms prove to perform much better than a standard SCGS algorithm View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Profiling Macro Data Flow Graphs for Parallel Implementation of FDTD Computations

    Page(s): 121 - 130
    Save to Project icon | Request Permissions | Click to expandQuick Abstract | PDF file iconPDF (381 KB) |  | HTML iconHTML  

    In this paper, we present methodology, which enables designing and profiling macro data flow graphs that represent computation and communication patterns for the FDTD (finite difference time domain) problem in irregular computational areas. Optimized macro dataflow graphs (MDFG) for FDTD computations are generated in three main phases: generation of initial MDFG based on wave propagation area partitioning, MDFG nodes merging with load balancing to obtain given number of macro nodes and communication optimization to minimize and balance internode data transmissions. The computation efficiency for several communication systems (MPI, RDMA RB, SHMEM) is discussed. Relations between communication optimization algorithms and overall FDTD computation efficiency are shown. Experimental results obtained by simulation are presented View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.