High-Performance Computing and Networking: 9th International Conference, HPCN Europe 2001, Amsterdam, The Netherlands, June 25-27, 2001, Proceedings (Lecture Notes in Computer Science, 2110)

This Book Constitutes The Refereed Proceedings Of The 9th International Conference On High-performance Computing And Networking, Hpcn Europe 2001, Held In Amsterdam, The Netherlands In June 2001. The 67 Revised Papers And 15 Posters Presented Were Carefully Reviewed And Selected From A Total Of Almost 200 Submissions. Among The Areas Covered Are Web/grid Applications Of Hpcn, End User Applications, Computational Science, Computer Science, And Java In Hpcn. Web- And Grid-based Applications -- A Virtual Data Grid For Ligo -- A Multidisciplinary Scientific Data Portal -- Migratable Sockets For Dynamic Load Balancing -- Toward Realizable Restricted Delegation In Computational Grids1 -- Profiling Facility On A Metasystem -- Utilizing Supercomputer Power From Your Desktop -- Tertiary Storage System For Index-based Retrieving Of Video Sequences -- Data And Metadata Collections For Scientific Applications -- The Vlam-g Abstract Machine: A Data And Process Handling System On The Grid -- Optimal Caching Policies For Web Objects -- A Computational Web Portal For The Distributed Marine Environment Forecast System -- Role Of Aging, Frequency, And Size In Web Cache Replacement Policies -- A Java-based Model Of Resource Sharing Among Independent Users On The Internet -- The Grb Library: Grid Computing With Globus In C -- Certificate Use For Supporting Merging And Splitting Of Computational Environments? -- Data Management For Grid Environments --^ Mobile Agents For Distributed And Dynamically Balanced Optimization Applications -- End User Applications -- A Comparison Of Three Parallel Algorithms For 3d Reconstruction -- Parallel 3d Adaptive Compressible Navier-stokes Solver In Geofem With Dynamic Load-balancing By Drama Library -- Parallel Models For Reactive Scattering Calculations -- The Use Of Intrinsic Properties Of Physical System For Derivation Of High-performance Computational Algorithms -- Parallel Dem Simulations Of Granular Materials -- A Generic Support For Distributed Deliberations -- Using Streaming And Parallelization Tecniques For 3d Visualization In A High Performance Computing And Networking Environment -- Performance Evaluation Of Parallel Groupby-before-join Query Processing In High Performance Database Systems -- Design And Implementation Of An Rpc-based Arc Kernel -- Computer Science -- Application-controlled Coherence Protocols For Scope Consistent Software Dsms --^ Source Code And Task Graphs In Program Optimization -- Event Manipulation For Nondeterministic Shared-memory Programs -- An Open Distributed Shared Memory System1 -- Two Layers Distributed Shared Memory -- Influence Of Compiler Optimizations On Value Prediction -- Experiments With Sequential Prefetching -- Code Positioning For Vliw Architectures -- Selective Register Renaming: A Compiler-driven Approach To Dynamic Register Renamin -- An Adaptive Space-sharing Policy For Heterogeneous Parallel Systems -- Orthogonal Processor Groups For Message-passing Programs -- Scheduling Task Graphs On Arbitrary Processor Architectures Considering Contention -- Pio: Parallel I/o System For Massively Parallel Processors -- Fast “short” Messages On A Linux Cluster -- Computational Science -- Efficient Monte Carlo Linear Solver With Chain Reduction And Optimization Using Plfg -- On-line Tool Support For Parallel Applications -- Cluster Computation For Flood Simulations1 --^ Advanced Library Support For Irregular And Out-of-core Parallel Computing -- A Parallel Adi Method For Linear And Non-linear Equations -- Study Of The Parallel Block One-sided Jacobi Method? -- A Practical Approach To Efficient Use Of Heterogeneous Pc Network For Parallel Mathematical Computation -- An Hierarchical Mpi Communication Model For The Parallelized Solution Of Multiple Integrals -- Impact Of Data Distribution On Performance Of Irregular Reductions On Multithreaded Architectures? -- Implementing And Benchmarking Derived Datatypes In Metacomputing -- Mpc++ Performance For Commodity Clustering -- Dynamic Instrumentation And Performance Prediction Of Application Execution -- Improving Load Balancing In A Parallel Cluster Environment Using Mobile Agents -- Posters -- Simulation And 3d Visualization Of Bioremediation Interventions In Polluted Soils -- Resource Planning In Converged Networks -- Parallelization Of The Stem-ii Air Quality Model --^ Simulating Parallel Architectures With Bsplab -- A Blocking Algorithm For Fft On Cache-based Processors -- Monte Carlo Simulations Of A Biaxial Liquid Crystal Model Using The Condor Processing System -- Generic Approach To The Design Of Simulation- Visualization Complexes -- On Parallel Programming Language Caper -- Increased Efficiency Of Parallel Calculations Of Fragment Molecular Orbitals By Using Fine-grained Parallelization On A Hitachi Sr8000 Supercomputer -- Customer-centered Models For Web-sites And Intra-nets -- A Prototype For A Distributed Image Retrieval System -- Data-object Oriented Design For Distributed Shared Memory1 -- Using Virtual User Account System For Managing Users Account In Polish National Cluster -- Performance Evaluation Of Xtp And Tcp Transport Protocols For Reliable Multicast Communications -- Parallelization Of The Continuous Global Optimization Problem With Inequality Constraints By Using Interval Arithmetic? -- Workshops --^ A Distributed Platform With Features For Supporting Dynamic And Mobile Resources -- Implementing An Efficient Java Interpreter -- Efficient Dispatch Of Java Interface Methods -- Implementation Of A Corba-based Metacomputing System? -- Joint: An Object Oriented Message Passing Interface For Parallel Programming In Java -- A Framework For Opportunistic Cluster Computing Using Javaspaces1) -- Scientific Computation With Javaspaces -- Computational Communities: A Marketplace For Federated Resources -- A Compiler Infrastructure For High-performance Java? -- Optimizing Java-specific Overheads: Java At The Speed Of C? -- Combining Batch And Streaming Paradigms For Metacomputing Applications -- Identification And Quantification Of Hotspots In Java Grande Programs -- Hepgrid2001: A Model Of A Virtual Data Grid Application -- Grid Services For Fast Retrieval On Large Multidimensional Databases. Bob Hertzberger, Alfons Hoekstra, Roy Williams (eds.). Includes Bibliographical References And Index. High-Performance Computing and Networking Preface Organization Table of Contents A Virtual Data Grid for LIGO 1 Physics 1.1 Data in LIGO 1.2 LIGO Data Objects 1.3 Computational Aspects, Pulsar Search 1.4 Event Identification Computation Pipeline 2 GriPhyN/LIGO Virtual Data Grid 2.1 Virtual Data 2.2 Simple Virtual Data Request Scenario 2.3 Generalized Virtual Data Description 3 GriPhyN Support 3.1 User Request 3.2 Data Access, Cost Performance Estimation 3.3 Proactive Data Replication 4 Conclusion References A Multidisciplinary Scientific Data Portal The Problem The CLRC Data Portal System Architecture User Requirement Capture SR Single Crystal Diffraction Scenarios ISIS Single Crystal Diffraction Scenario User Interface Metadata Model and Implementation Data Resource Interfaces Project Status and Future Plans References Migratable Sockets for Dynamic Load Balancing Introduction Requirements for Socket Migration Description of the msocket System Redirection of TCP Connections Implementation Tests of Functionality and Overhead Conclusions and Future Work References Toward Realizable Restricted Delegation in Computational Grids 1 Introduction 2 Legion Security Overview 3 Delegation Approaches 3.1 Method Restrictions 3.2 Object Restrictions 3.3 Time-Dependent Restrictions 4 Related Work 5 Conclusions References Profiling Facility on a Metasystem Introduction Profiling Profiling on a Metacomputer Profile Variation Profiling Levels Condensed Profiles Overhead on the Metasystem Support for Profiling in SUMA Description of SUMA A Model for Profiling in SUMA Experiments Conclusions References Utilizing Supercomputer Power from Your Desktop Introduction Requirements Design Implementation SPINEware Object Servers Web-Interface Server Web-Interface Java Applets A Case Study: Superbroker Conclusions References Tertiary Storage System for Index-Based Retrieving of Video Sequences Introduction Related Works Video Tertiary Storage System Repository Daemon Tertiary File Manager Daemon VTSS and MMSRS Performance Comparison Multimedia Storage and Retrieval System Test Results Conclusions and Future Work References Data and Metadata Collections for Scientific Applications Introduction Data Handling Infrastructure Application Summary References The VLAM-G Abstract Machine: A Data and Process Handling System on the Grid Introduction Experimenting in VL VL-AM Architecture VL-AM Front End VL-AM RTS Design Implementation Related Work Conclusions References Optimal Caching Policies for Web Objects Introduction The Model Dynamic Programming Equations Numerical Results Conclusion References A Computational Web Portal for the Distributed Marine Environment Forecast System Introduction Distributed Marine Environment Forecast System Support for Complex Multistep Computational Tasks Platform Independence Persistence Middle Tier Application Server Web Access Security Summary References Role of Aging, Frequency, and Size in Web Cache Replacement Policies Introduction Related Work and Background GDFS Cache Replacement Policy: Formal Definition Data Collection Sites Basic Simulation Results Generalized GDFS Policy and Its Simulation Results Conclusion and Future Work References A Java-Based Model of Resource Sharing among Independent Users on the Internet 1 Resource Sharing among Autonomous Users on the Internet 2 A Java Model of Resource Sharing 3 Protocols for Sharing 3.1 Location of Available Resources 3.2 Resource Acquisitation by Clients 3.3 Disk Storage Sharing 4 Implementing Resource Sharing in the Jini Programming Environment References The GRB Library: Grid Computing with Globus in C 1 Introduction 2 Library Design 3 Implementation 4 The Grid Resource Broker 5 Related Work and Conclusions References Certificate Use for Supporting Merging and Splitting of Computational Environments Introduction Using X.509 Certificates The Use of Certificate Revocation Lists Conclusions and Future Work References Data Management for Grid Environments Foundations Data Management - A Wider Perspective Support for Data Storage and Access The Minimum Unit and Access Patterns Support for Metadata Management Standards Support and Application Services Supporting Primitive and Global Services Conclusion References Mobile Agents for Distributed and Dynamically Balanced Optimization Applications Introduction The Branch and Bound Technique Extensions to the Aglet Mobile Agent System Extension of the Multicast Primitive The Dynamic Workload Balancing Service The Distributed Mobile-Agent Based B&B Experimental Results Conclusions References A Comparison of Three Parallel Algorithms for 3D Reconstruction Introduction and Motivation Basic Concepts: Algorithms and Notations Performance Data Conclusions References Parallel 3D Adaptive Compressible Navier-Stokes Solver in GeoFEM with Dynamic Load-Balancing by DRAMA Library 1 Introduction 2 Parallel 3D Compressible Navier-Stokes Solver: epHYBRID 3 Parallel Mesh Adaptations and DYnamic Load-Balancing Module: pADAPT/DRAMA 3.1 pADAPT 3.2 DRAMA and Data Migration 4 Distributed Data Structures for Parallel Mesh Adaptation 5 Numerical Examples 5.1 Parallel Performance of epHYBRID without Adaption 5.2 Comparison of Repartitioning Methods 6 Conclusion References Parallel Models for Reactive Scattering Calculations Introduction Quantum Approaches to Reactive Probabilities, Cross Sections, and Rate Coefficients The Time Independent Computational Procedure The Parallelization of the Time Independent Code The Time Dependent Computational Procedure The Parallelization of the Time Dependent Code Conclusions References The Use of Intrinsic Properties of Physical System for Derivation of High-Performance Computational Algorithms 1 Introduction 2 The Main Idea of Approach 3 The Equations of Nonequilibrium Kinetics 4 Inelastic Scattering Amplitude 5 Concluions References Parallel DEM Simulations of Granular Materials Introduction The Distinct Element Method Contact Detection Numerical Contact Models The DEM Algorithm The Parallel DEM Algorithm Visualization Applications in Granular Media Simulations Rockfall Impact on an Embankment Powder Packings Conclusion References A Generic Support for Distributed Deliberations 1 Introduction 2 A Demonstrative Example 3 Main Design Choices 3.1 A Meta Model for Describing Deliberation Models 3.2 Dynamic Deliberation Model Refinement 3.3 Argumentation and Decision Support 3.4 Synchronous and Asynchronous Work Support 4 DOTS Prototype 4.1 DOTS Architecture and Tools 4.2 DOTS Usage 5 Conclusion References Using Streaming and Parallelization Techniques for 3D Visualization in a High-Performance Computing and Networking Environment 1 Introduction and Motivation 2 Requirements 3 Concept and Implementation 3.1 Efficient Creation of 3D Objects 3.2 3D Streaming System 4 Experiences from an Example Application 5 Conclusion and Future Work References Performance Evaluation of Parallel GroupBy-Before-Join Query Processing in High Performance Database Systems 1 Introduction 2 GroupBy Queries: A Background 3 Parallel Algorithms for “GroupBy-Before-Join” Query 3.1 Early Distribution Scheme 3.2 Early GroupBy with Partitioning Scheme 3.3 Early GroupBy with Replication Scheme 4 Performance Evaluation 4.1 Result of Experiment 4.2 Discussions 5 Conclusions and Future Work References Design and Implementation of an RPC-Based ARC Kernel Introduction Architecture of the RPC-Based ARC Kernel Resource Allocation Protocol Dynamic Join/Leave Protocol Horse Power Factor Dynamic Task Distribution Performance and Overheads Conclusion References Application-Controlled Coherence Protocols for Scope Consistent Software DSMs 1 Introduction 2 Memory Coherence Protocols for DSM Systems 3 The JIAJIA Software DSM System 4 The Proposed Cache Coherence Mechanism 5 Experimental Results 6 Related Works 7 Conclusions and Future Work References Source Code and Task Graphs in Program Optimization Introduction Basic Definitions Program Information and Task Graphs The Optimization Algorithms Redundant Scheduling Diamond Scheduling Practical Results Related Work Conclusion References Event Manipulation for Nondeterministic Shared-Memory Programs Introduction The Event Graph - A Parallel Program Model Nondeterminism in Shared-Memory Programs Event Manipulation Approach Example Conclusions and Future Work References An Open Distributed Shared Memory System 1 Introduction 2 System Overview 3 Experimental Results 3.1 Use of the Open DSM System to Resolve the ASP Problem 3.2 Performance of Read - Write Operations in Different DSM Systems Constructed from the Open DSM System 4 Conclusion and Future Work References Two Layers Distributed Shared Memory Introduction DVSA SHOB SHOB Primitives SHOB Implementation Experimental Results The Performance of a Uniform Multigrid Method References Influence of Compiler Optimizations on Value Prediction Introduction Related Work Evaluation Methodology Processor Model Benchmark Binaries Simulation Results Predictability Processor Performance Conclusion References Experiments with Sequential Prefetching Introduction The Analysis Basis One-Line-Lookahead Speedup Due to Prefetching Two-Line-Lookahead Multi-Line-Lookahead: Generalized Analysis Practical Limitations Experimental Results Summary and Conclusions References Code Positioning for VLIW Architectures Introduction Related Work Code Positioning Algorithm Algorithm Description Evaluation Experimental Setup Experimental Results Conclusions References Selective Register Renaming: A Compiler-Driven Approach to Dynamic Register Renaming Introduction Background Selective Register Renaming Experiments Evaluation of Register Utilization Evaluation of Instruction Level Parallelism Conclusions and Future Work: The Precise Interrupts References An Adaptive Space-Sharing Policy for Heterogeneous Parallel Systems Introduction Space-Sharing Policies for Homogeneous Systems Adaptive Policy AP Modified Adaptive Policy MAP Proposed Policy for Heterogeneous Systems Job and Workload Characterization Job Structure The Workload Model Simulation Model Performance Comparison AP versus HAP Performance Improvement of HAP Over MAP Sensitivity to Parameter {em f} Summary References Orthogonal Processor Groups for Message-Passing Programs Introduction Orthogonal Structures of Computations in Scientific Applications Mapping to Orthogonal Processor Groups Runtime Tests on Cray T3E Comparison to Related Work Conclusions and Future Work References Scheduling Task Graphs on Arbitrary Processor Architectures Considering Contention Introduction Scheduling Algorithms Dynamic Level Scheduling (DLS) Bubble Scheduling and Allocation (BSA) The Proposed Algorithm Experimental Comparison Conclusion References PIO: Parallel I/O System for Massively Parallel Processors Introduction Experimental Environment Parallel I/O System Model of PIO System API Communication Load Balancing Supporting Multiple Platforms Performance Evaluation Basic Data Transfer Dynamic Load Balancing PAVEMENT -- A PIO Application Conclusions and Future Works References Fast ``Short'' Messages on a Linux Cluster Introduction Communication Primitives Communication Primitives in Kernel Space Implementation Details Experimental Results Conclusions References Efficient Monte Carlo Linear Solver with Chain Reduction and Optimization Using PLFG Introduction Stochastic Methods for Solving SLAE Markov Process Minimal Probable Error Almost Optimal Transition Frequency Function Parameters Estimation Lagged Fibonacci Generators The {tt PLFG} Parallel Pseudo-random Number Generator Numerical Tests Conclusion References On-Line Tool Support for Parallel Applications Introduction OMIS as Basis for Building a Tool Environment Operation of PATOP on Top of the OCM for MPI Gathering Performance Data with the OCM Adaptation of the Tool Environment to MPI Monitoring Overhead Test Interoperability within the Tool Environment Concluding Remarks References Cluster Computation for Flood Simulations 1 Introduction 2 Mathematical Models and Motivations 3 Numerical Solutions 3.1 Recursive Boubling 3.2 Cyclic Reduction 3.3 Comparison of Methods 4 Parallel Implementations 4.1 A Linux Cluster and the SGI Origin2000 4.2 Parallization Method 4.3 Experiments and Measurements 5 Conclusion and Future Work References Advanced Library Support for Irregular and Out-of-Core Parallel Computing Introduction Design Goals Library Structure and Application Interface Data Storage, Communication, Computation, and I/O Library Design Application Interface Performance Evaluation texttt {lip} vs. Virtual Memory Out-Of-Core Code Coupled to the Hilbert Partitioner Parallelisation of the AVL FIRE Benchmark Summary References A Parallel ADI Method for Linear and Non-linear Equations Introduction General Formulation and Numerical Approximation Algorithm Parallel Scheme Conclusion References Study of the Parallel Block One-Sided Jacobi Method Introduction Sequential Algorithm Study of the Parallelization Overlapping the Communications Experimental Results Block Version Parallelization Overlapping the Communications Experimental Results Conclusion References A Practical Approach to Efficient Use of Heterogeneous PC Network for Parallel Mathematical Computation 1 Introduction 2 ScaLAPACK on Homogeneous Clusters 3 The Programming envIronment for Network of COmputers (PINCO) 3.1 Evaluating System Resources 3.2 Resource and Task Manager 3.3 Application Program Interface 4 PINCO Application to ScaLAPACK Library 4.1 ScaLAPACK Data Distribution 4.2 The Experiments Perfomed on Solving Generalized Eigenproblems and on LU Factorization Routines 5 Conclusions References An Hierarchical MPI Communication Model for the Parallelized Solution of Multiple Integrals Introduction Physical and Mathematical Background A Motivating Example from Polymer Physics The Seven-Point Simpson Rule Recursively Determined Adaptive Step-Widths Parallelization and Communication Model Simple Parallelization Model Complex Parallelization Model with Communicator Trees Results and Discussion Parallel Efficiency Communication Rates Calculation Speed Summary and Conclusions References Impact of Data Distribution on Performance of Irregular Reductions on Multithreaded Architectures Introduction Execution Strategy Irregular Reduction Loops Execution Strategy for Multithreaded Architectures Experimental Results Multithreaded Architecture Benchmarks and Experiment Design Results Summary References Implementing and Benchmarking Derived Datatypes in Metacomputing Introduction Concept of PACX-MPI Analysis of MPI's Derived Datatypes Pack/Unpack Model Shared Memory Model Direct Communication Model A Grid-Enabled Implementation of Derived Datatypes The Implementation of Derived Datatypes in PACX-MPI Tests and Results Results on the Cray T3E Results on the NEC SX5 Summary References MPC++ Performance for Commodity Clustering Introduction MPC++ and Port to MPI Evaluation Procedure Porting NPB CG and IS to MPC++ Performance Evaluation Evaluation Environment Benchmark Results of CG and IS Details of CG (Class A) Conclusion References Dynamic Instrumentation and Performance Prediction of Application Execution 1 Introduction 2 PACE: Performance Analysis and Characterisation Environment 2.1 Model Components 2.2 Model Creation 3 Dynamic Instrumentation of Application Execution 3.1 Overview 3.2 An Automated Profiling Technique for PACE 3.3 Source Code Instrumentation and Dynamic Analysis 4 Analysis and Conclusion References References Improving Load Balancing in a Parallel Cluster Environment Using Mobile Agents Introduction Cluster Environment Mobile Agents Systems Experimental Results Introduction Experiments Conclusion References Simulation and 3D Visualization of Bioremediation Interventions in Polluted Soils 1 Introduction 2 The Bioremediation Simulation 3 The 3D-Visualization Application 3.1 The Visualization Methods References Resource Planning in Converged Networks 1 The Resource Planning Model 2 Implementation and Test of the Model Conclusion References Parallelization of the STEM-II Air Quality Model Introduction Sequential Code Analysis Parallelization of the Code Experimental Results Conclusions References Simulating Parallel Architectures with BSPlab 1 Introduction 2 A Short Introduction to BSP and BSPlab 2.1 Architectural Models in BSPlab 3 Status and Example Experiment 4 Concluding Remarks References A Blocking Algorithm for FFT on Cache-Based Processors Introduction A Block Six-Step FFT Algorithm Performance Results Conclusion References Monte Carlo Simulations of a Biaxial Liquid Crystal Model Using the Condor Processing System 1 Introduction 2 The Biaxial Liquid Crystal Lattice Model 3 The CONDOR Processing System 4 LC Simulations Using the INFN CONDOR Pool References Generic Approach to the Design of Simulation-Visualization Complexes 1 Introduction 2 Simulation-Visualization Processes 3 Interaction between Simulation and Visualization Compounds 4 Conclusion References On Parallel Programming Language Caper 1 Introduction 2 Program Structure, Data, and Control 3 Caper Virtual Machine 4 Programming Experience in Caper References Increased Efficiency of Parallel Calculations of Fragment Molecular Orbitals by Using Fine-Grained Parallelization on a HITACHI SR8000 Supercomputer Introduction Results and Discussion Conclusions References Customer-Centered Models for Web-Sites and Intra-Nets Practical Experience with Internet Event Mining Heaps' and Zipf's Laws Internet Event Mining Prerequisites The Markov Models Applications Modeling without Markov Processes n-Markov Chains and Further Discussion of Markov Models Building Innovative Retrieval Tools Conclusions References A Prototype for a Distributed Image Retrieval System Introduction Parallel Components Conclusions References Data-Object Oriented Design for Distributed Shared Memory 1 Introduction 2 Basic Concepts 3 Implementation Issues References Using Virtual User Account System for Managing Users Account 1 Introduction 2 Polish National Cluster 3 Virtual User Account System 4 Cluster Implementation 5 Conclusion References Performance Evaluation of XTP and TCP Transport Protocols for Reliable Multicast Communications Introduction Cluster Environment and XTP Protocol Experimental Results Conclusion References Parallelization of the Continuous Global Optimization Problem with Inequality Constraints by Using Interval Arithmetic Introduction Hansen's Sequential Algorithm First Parallel Version Parallel Version with the Load Balancing Strategy Conclusions References A Distributed Platform with Features for Supporting Dynamic and Mobile Resources Introduction Related Work Network Centric Approach Computation Centric Approach Other Approaches -- Global Computing The Jem Platform Main Goals of the Platform The Platform Jem for Global Computing Support for Dynamism and Mobility Smart Proxies A Dynamic Naming Service Comparison to Related Work Conclusion References Implementing an Efficient Java Interpreter Introduction Virtual Machine Interpreters Implementing a Java Interpreter The Byte Code Translator Optimizations Instruction Definition Interpreter Generator Preliminary Experimental Results Conclusion References Efficient Dispatch of Java Interface Methods Introduction Virtual Method Dispatch in Jalape~{n}o Interface Method Tables Performance Related Work Conclusion References Implementation of a CORBA-Based Metacomputing System Introduction CORBA Specification of SUMA General Description Execution Agents Client Stub Proxy Scheduler User Control Execution Environment Experiments Related Work Conclusions and Future Work References JOINT: An Object Oriented Message Passing Interface for Parallel Programming in Java Introduction The JOINT Programming Interface AppTask Tasker Message Communicator Example: Searching for Prime Numbers Implementation Status Conclusions and Future Work References A Framework for Opportunistic Cluster Computing Using JavaSpaces 1 Introduction 2 A Framework for Opportunistic Parallel Computing on Clusters 2.1 Framework Architecture 2.2 Implementation and Operation 3 Framework Evaluation 3.1 Scalability Analysis 3.2 Adaptation Protocol Analysis 3.3 Dynamic Behavior Patterns under Varying Load Conditions 4 Conclusions References Scientific Computation with JavaSpaces Introduction Low-Level Metrics Parallel Algorithms Parametric Experiments Dynamic Compilation Conclusions References Computational Communities: A Marketplace for Federated Resources Introduction A Computational Economy Overview Building a Computational Community Resource Discovery and Selection Computational Supply and Demand Implementation Related Work Conclusions and Future Work References A Compiler Infrastructure for High-Performance Java Introduction Overview Architecture Robust Framework Symbol Table Representation External Class Resolution Modularity of Analyses Implementation Details zJava Compiler Front-End Collections HLIR Classes Examples Statement Insertion Data Structure Visualization Software Architecture Visualization Related Work Concluding Remarks References Optimizing Java-Specific Overheads: Java at the Speed of C? Introduction The Manta Compiler Closed-World Assumption Object Inlining Method Inlining Escape Analysis Array Bounds-Check Deactivation and Elimination Application Kernel Performance Iterative Deepening A* (IDA*) Traveling Salesperson Problem (TSP) Successive Overrelaxation (SOR) Conclusions References Combining Batch and Streaming Paradigms for Metacomputing Applications Introduction Combining Batch and Streaming Paradigms Metacomputing in Amica The Streaming Concept in Amica Implementation of the Streaming Concept An Example Related Work Conclusions and Future Work References Identification and Quantification of Hotspots in Java Grande Programs Introduction Grande Programs Measured Dynamic Method Execution Frequencies Conclusions References HEPGRID2001: A Model of a Virtual Data Grid Application Introduction Hardware Model Data Model Workload Model Job Model Properties of Physics Analysis Workloads Workload Details Workload Statistics Conclusions References Grid Services for Fast Retrieval on Large Multidimensional Databases 1 Introduction 2 The ESTEDI Project 3 The RasDaMan Array DBMS 3.1 Modelling and Querying Array Data in RasDaMan 3.2 Application Development and Client/Server Communication 3.3 Physical Array Storage and Processing 4 Conclusion References Appendix Author Index