"A record-setting attendance was anticipated for ISC-HPC 2020 in Frankfurt, but as with all other conferences in summer 2020, theglobal coronavirus pandemic forced it to be a digital event."
Includes author index.
Intro -- Preface -- Organization -- Contents -- Architectures, Networks and Infrastructure -- FASTHash: FPGA-Based High Throughput Parallel Hash Table -- 1 Introduction -- 2 Related Work -- 2.1 Hash Table Implementation on CPU and GPU -- 2.2 Hash Table Implementation on FPGA -- 2.3 Novelty of Our Work -- 3 Hash Table Overview -- 3.1 Definition of Hash Table -- 3.2 Parallel Hash Table -- 4 FASTHash: An FPGA-Based Parallel Hash Table -- 4.1 Hash Table Data Organization -- 4.2 Hash Table Architecture -- 4.3 Customization for Static Hash Table -- 5 Hash Table Guarantees and Applications Supported
5.1 Implications of Relaxed Eventual Consistency -- 5.2 Applications Supported -- 6 Experiments and Results -- 6.1 Experimental Methodology -- 6.2 Results -- 6.3 Comparison with State-of-the-Art (SOTA) Designs -- 7 Conclusion -- References -- Running a Pre-exascale, Geographically Distributed, Multi-cloud Scientific Simulation -- 1 Introduction -- 1.1 Related Work -- 2 The Workload Management System Setup -- 2.1 The Multi-cloud, Geographically Distributed HTCondor Setup -- 2.2 Dealing with Data Handling -- 2.3 Unexpected Problems Encountered in the HTCondor Setup
3 The Multi-cloud, Multi-region Setup -- 3.1 The Social Hurdle -- 3.2 Provisioning the 51k GPUs Over 3 Cloud Providers Using Multiple Regions -- 3.3 An Overview of the Provisioned Resources -- 3.4 Preparations -- 3.5 Cloud Cost Analysis -- 4 The IceCube Science Proposition -- 4.1 The IceCube Neutrino Observatory -- 4.2 The Importance of Proper Calibration -- 4.3 Using GPUs for Photon Propagation Simulation -- 4.4 The Science Output -- 5 Conclusions -- References -- Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)TM Streaming-Aggregation Hardware Design and Evaluation
1 Introduction -- 2 Previous Work -- 3 Streaming-Aggregation -- 3.1 Tree Type -- 3.2 InfiniBand Transport Selection -- 3.3 Tree Locking -- 3.4 Reduction Tree -- 3.5 Reduction Pipelining -- 3.6 Switch-Level Reduction -- 3.7 Result Distribution -- 3.8 Aggregation Protocol Resilience -- 4 Experiments -- 4.1 Test System Configuration -- 4.2 Synthetic Benchmarks -- 4.3 Application Benchmarks -- 5 Summary -- References -- Artificial Intelligence and Machine Learning -- Predicting Job Power Consumption Based on RJMS Submission Data in HPC Systems -- 1 Introduction -- 1.1 Constraints for Job Scheduling
1.2 Related Work -- 1.3 Contributions -- 2 Extracted Data and Preprocessing -- 2.1 The COBALT Supercomputer and The SLURM RJMS -- 2.2 From Raw Data to Relevant Features -- 2.3 Target and Problem Formalization -- 3 Instance Based Regression Model -- 3.1 Inputs as Categorical Data -- 3.2 An Input-Conditioning Model -- 3.3 Variable Selection -- 4 Global Consumption Practical Estimation -- 4.1 Weighted Estimator for Global Power Estimation -- 4.2 Online Computations -- 4.3 Exponential Smoothing for Weighted and Streamed Update -- 5 Numerical Results and Discussion -- 5.1 Offline Instance-Based Model -- 5.2 Comparison with the Offline IBmodel