Network and Parallel Computing: 13th IFIP WG 10.3 International Conference, NPC 2016, Xi'an, China, October 28-29, 2016, Proceedings (Lecture Notes in Computer Science Book 9966)

This book constitutes the proceedings of the 13th IFIP WG 10.3 International Conference on Network and Parallel Computing, NPC 2016, held in Xi'an, China, in October 2016. The 17 full papers presented were carefully reviewed and selected from 99 submissions. They are organized in the following topical sections; memory: non-volatile, solid state drives, hybrid systems; resilience and reliability; scheduling and load-balancing; heterogeneous systems; data processing and big data; and algorithms and computational models. Preface 6 Organization 7 Contents 9 Memory: Non-Volatile, Solid State Drives, Hybrid Systems 11 VIOS: A Variation-Aware I/O Scheduler for Flash-Based Storage Systems 12 1 Introduction 12 2 Background and Related Work 14 2.1 SSD Organization 14 2.2 Process Variation of Flash Memory 14 2.3 I/O Scheduler for Flash-Based SSDs 15 3 Details of VIOS 16 3.1 Block Management 16 3.2 Global Chip-State Vector 17 3.3 Conflict Optimized Scheduling Mechanism 17 4 Experimental Results 19 4.1 Performance Analysis of VIOS 20 4.2 Sensitivity Analysis of VIOS 22 5 Conclusion 24 References 24 Exploiting Cross-Layer Hotness Identification to Improve Flash Memory System Performance 26 1 Introduction 26 2 Background and Related Work 28 2.1 Tradeoff Between Flash Cell Wearing and Read Latency 28 2.2 Related Work 29 3 Exploiting Cross-Layer Hotness Identification to Improve Read and Endurance Performance (HIRE) 30 3.1 Cross-Layer Study for Hotness Identifier 30 3.2 Voltage Controller in HIRE 31 3.3 Overhead Analysis 32 4 Performance Evaluation 33 4.1 Methodology 33 4.2 Experiment Results 34 5 Conclusion 35 References 36 Efficient Management for Hybrid Memory in Managed Language Runtime 38 1 Introduction 39 2 Managed Runtime Background 40 3 Related Work 41 4 Hybrid Memory Management Scheme 42 4.1 Overview 42 4.2 HMprof Offline Performance Instrumentation 43 4.3 Hot Object Marking 45 5 Experimental Methodology 45 6 Evaluation Results 46 6.1 Heap Partition Placement 46 6.2 Hot Object Allocation 48 7 Conclusion 49 References 49 Resilience and Reliability 52 Application-Based Coarse-Grained Incremental Checkpointing Based on Non-volatile Memory 53 Abstract 53 1 Introduction 53 2 Background and Motivation 55 2.1 Non-volatile Memory 55 2.2 Incremental Checkpointing with Non-volatile Memory 55 2.3 The Problem with Current Checkpoint 56 3 Design and Implementation 57 3.1 Contiguous Memory Regions to be Visited 57 3.2 Application-Based Coarse-Grained Checkpoint: Loose Monitoring Granularity for “Hot” Applications 59 3.3 Huge Page Support 60 4 Experiments 61 4.1 Experimental Setup and Benchmarks 61 4.2 Performance Metrics and Corresponding Results 61 5 Related Work 64 6 Conclusion 64 Acknowledgments 64 References 65 DASM: A Dynamic Adaptive Forward Assembly Area Method to Accelerate Restore Speed for Deduplication-Based Backup Systems 66 1 Introduction 66 2 Background and Motivation 67 2.1 Data Deduplication 68 2.2 Fragmentation Problem and Restoration Speed 68 2.3 Forward Assembly Area 69 2.4 Our Motivation 70 3 Design of DASM 70 3.1 Dynamic Orthotics 72 3.2 Near-Optimal Cache Policy 73 4 Performance Evaluation 73 4.1 Evaluation Methodology 74 4.2 Experimental Results 74 4.3 Analysis 76 5 Conclusions 77 References 77 Scheduling and Load-Balancing 79 A Statistics Based Prediction Method for Rendering Application 80 1 Introduction 80 2 Related Works 81 3 Proposed Strategy 82 3.1 System Design 82 3.2 Parsing of Parameters 83 3.3 Strategy Description 86 4 Experimental Results 86 5 Conclusion and Future Work 89 References 89 IBB: Improved K-Resource Aware Backfill Balanced Scheduling for HTCondor 92 Abstract 92 1 Introduction 92 2 Related Work 93 3 Improved Backfill Balance Scheduling 93 3.1 Thought of IBB 93 3.2 Apply Backfilling to HTCondor 96 4 Experiments and Evaluations 96 4.1 Simulation Experiments 96 4.2 Experiments Conducted on HTCondor 97 4.3 Discussion 98 5 Conclusion and Future Work 98 Acknowledgments 99 References 99 Multipath Load Balancing in SDN/OSPF Hybrid Network 100 Abstract 100 1 Introduction 100 2 Problem Formulation 101 2.1 Hybrid Network Scenario 101 2.2 Optimization Problem Definition 102 3 Load Balancing in Hybrid SDN Network 103 3.1 Disjoint Multipath Calculation 103 3.2 LRU: A New FPTAS Algorithm 103 4 Experiments and Evaluation 105 5 Conclusion 106 Acknowledgments 106 References 107 Heterogeneous Systems 108 A Study of Overflow Vulnerabilities on GPUs 109 1 Introduction 109 2 Background on CUDA Architecture 111 3 Empirical Evaluation of GPU Vulnerabilities 111 3.1 Experiment Setup 112 3.2 Stack Overflow 112 3.3 Heap Overflow 114 3.4 Other Vulnerabilities 117 4 Discussions and Countermeasures 118 5 Related Work 119 6 Conclusion 120 References 120 Streaming Applications on Heterogeneous Platforms 122 1 Introduction 122 2 Related Work 123 3 A Statistical View 125 3.1 Benchmarks and Datasets 125 3.2 Experimental Platforms 125 3.3 Measurement Methodology 126 3.4 Results and Analysis 126 4 Our Streaming Approach 128 4.1 Categorization 128 4.2 Code Streaming 129 5 Experimental Results 132 6 Conclusion 133 References 134 Data Processing and Big Data 136 DSS: A Scalable and Efficient Stratified Sampling Algorithm for Large-Scale Datasets 137 1 Introduction 137 2 Related Work 138 3 Stratified Sampling Queries 139 3.1 A Single Stratified Sampling Query 139 3.2 Sequential Answering Process 140 4 Distributed Sampling Design 141 4.1 Sampling Representativeness in Distributed Environment 141 4.2 Distributed Algorithm Spark-SQE 141 4.3 Improved Distributed Algorithm DSS 143 5 Experimental Evaluation 145 5.1 Experimental Setup 145 5.2 Results 146 6 Conclusion 149 References 149 A Fast and Better Hybrid Recommender System Based on Spark 151 1 Introduction 151 1.1 Hybrid Recommender Systems 152 1.2 Work of Paper 152 2 Related Work 152 2.1 Recommender Systems 153 2.2 Weight Analysis 154 2.3 Spark 154 3 Design Overview 155 3.1 Objective Function 155 3.2 Weight Calculation 156 4 Implementation 157 4.1 Modules 157 4.2 Discussion 158 5 Performance 158 5.1 Evaluation Index 158 5.2 Experimental Setup 159 5.3 Performance Comparision 159 6 Conclusion and Future Work 161 References 162 Discovering Trip Patterns from Incomplete Passenger Trajectories for Inter-zonal Bus Line Planning 164 Abstract 164 1 Introduction 164 2 Related Work 165 3 Problem Definition 167 4 Mining Frequent Trip Patterns for Inter-zonal Bus Lines 168 4.1 Mining Commuters 168 4.2 Mining FBPT4BL Patterns 170 4.3 Time Complexity Analysis 172 5 Evaluation 172 5.1 Sensitivity Analysis of MinPts and ɛ 172 5.2 Effective and Efficient Analysis of MCFTP Approach 173 6 Conclusion 175 Acknowledgments 175 References 175 FCM: A Fine-Grained Crowdsourcing Model Based on Ontology in Crowd-Sensing 176 Abstract 176 1 Introduction 176 2 Characteristic Ontology Triple 177 3 Recommendation Method 178 3.1 Recommendation Index 178 3.2 Service Decision Tree 179 4 Experiment 180 5 Conclusion 182 Acknowledgments 182 References 183 QIM: Quantifying Hyperparameter Importance for Deep Learning 184 1 Introduction 184 2 Background 185 2.1 Deep Learning (DL) 185 2.2 PB Design 186 3 QIM 187 3.1 Overview 187 3.2 Identifying the Value Range for Each Hyperparameter 187 3.3 QIM 188 4 Experimental Setup 188 5 Evaluation 188 5.1 Supervised Learning 188 5.2 Unsupervised Pre-training 190 5.3 Time Cost 190 6 Related Work 191 7 Conclusion 191 References 191 Algorithms and Computational Models 193 Toward a Parallel Turing Machine Model 194 1 Introduction and Motivation 194 2 Existing Work on Parallel Turing Machine --- A Disappointing Status Report 196 2.1 Existing Parallel Turing Machine Proposals 196 2.2 The Disappointing Status on Parallel Turing Machine Studies 197 3 A Parallel Turing Machine Model --- Our Proposal 198 3.1 The Concept of Codelets and Codelet Graphs (CDGs) 198 3.2 The Parallel Turing Machine Model 199 3.3 An Example to Illustrate How the PTM Works 201 3.4 Determinacy Property of Our Proposed PTM 203 4 Related Work 203 4.1 Parallel Turing Machine 203 4.2 Memory Consistency Models 204 4.3 The Codelet Model 204 4.4 Work on Parallel Computation Models 205 5 Conclusion and Future Work 205 References 205 On Determination of Balance Ratio for Some Tree Structures 208 1 Introduction 208 2 The Basic Properties and Algorithms 208 3 The Improvement of Time Complexity 212 4 Concluding Remarks 215 References 215 Author Index 216 Front Matter....Pages I-X Front Matter....Pages 1-1 VIOS: A Variation-Aware I/O Scheduler for Flash-Based Storage Systems....Pages 3-16 Exploiting Cross-Layer Hotness Identification to Improve Flash Memory System Performance....Pages 17-28 Efficient Management for Hybrid Memory in Managed Language Runtime....Pages 29-42 Front Matter....Pages 43-43 Application-Based Coarse-Grained Incremental Checkpointing Based on Non-volatile Memory....Pages 45-57 DASM: A Dynamic Adaptive Forward Assembly Area Method to Accelerate Restore Speed for Deduplication-Based Backup Systems....Pages 58-70 Front Matter....Pages 71-71 A Statistics Based Prediction Method for Rendering Application....Pages 73-84 IBB: Improved K-Resource Aware Backfill Balanced Scheduling for HTCondor....Pages 85-92 Multipath Load Balancing in SDN/OSPF Hybrid Network....Pages 93-100 Front Matter....Pages 101-101 A Study of Overflow Vulnerabilities on GPUs....Pages 103-115 Streaming Applications on Heterogeneous Platforms....Pages 116-129 Front Matter....Pages 131-131 DSS: A Scalable and Efficient Stratified Sampling Algorithm for Large-Scale Datasets....Pages 133-146 A Fast and Better Hybrid Recommender System Based on Spark....Pages 147-159 Discovering Trip Patterns from Incomplete Passenger Trajectories for Inter-zonal Bus Line Planning....Pages 160-171 FCM: A Fine-Grained Crowdsourcing Model Based on Ontology in Crowd-Sensing....Pages 172-179 QIM: Quantifying Hyperparameter Importance for Deep Learning....Pages 180-188 Front Matter....Pages 189-189 Toward a Parallel Turing Machine Model....Pages 191-204 On Determination of Balance Ratio for Some Tree Structures....Pages 205-212 Back Matter....Pages 213-213